1 INFO-VAX	Fri, 18 Apr 2003	Volume 2003 : Issue 213       Contents:% *&& << 1GB >> Flash Memory Drive (&(*  Re: 7.1-1 H1 to 7.3: issue?  Average Uptimes by OS  Re: Average Uptimes by OS  Re: Average Uptimes by OS  Re: Average Uptimes by OS  Re: Average Uptimes by OS  Re: Average Uptimes by OS : Cache Fusion? (Was: Re: Excellent new OpenVMS Testimonial)9 RE: DCL question (of the day)- null byte in string symbol 9 Re: DCL question (of the day)- null byte in string symbol 9 RE: DCL question (of the day)- null byte in string symbol E Help! Bizzare behaviour of C RTL function stat - issue for SAMBA port I Re: Help! Bizzare behaviour of C RTL function stat - issue for SAMBA port I Re: Help! Bizzare behaviour of C RTL function stat - issue for SAMBA port 6 How to access DB2 on an AS/400 from Oracle on OpenVMS? Re: HP Shipping Charges  Re: HTML favourite editor? Interesting VMS ad in Google* RE: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?* Re: Mirroring versus Shadowing - any diff?E Re: Personal Firewalls - was Email to Geoff.Graves@hp.com is rejected ) Problem with formfeed on the Genicom 4400 - Re: Problem with formfeed on the Genicom 4400 - Re: Problem with formfeed on the Genicom 4400 - Re: Problem with formfeed on the Genicom 4400 - Re: Problem with formfeed on the Genicom 4400  Re: SARS Impact on HP  Re: SARS Impact on HP  Re: StorageTek on SAN with VMS Strange SYS$ENQ problem  Re: Strange SYS$ENQ problem  Re: Strange SYS$ENQ problem  system tools cd mount problem & tcpip 5.3-1 socket inactivity timeout?& Re: TCPIP Services and IP masquerading; Re: Truly effective - let's get them for the VMS campaign!! ; Re: Truly effective - let's get them for the VMS campaign!!   Re: VAX C++ 5.6 install question  Re: VAX C++ 5.6 install question  Re: VAX C++ 5.6 install question  Re: VAX C++ 5.6 install question Re: VAX/Alpha VMS differences  Re: VAX/Alpha VMS differences  Re: VAX/Alpha VMS differences 5 Re: VMS to be present at european security conference 5 Re: VMS to be present at european security conference 5 Re: VMS to be present at european security conference  Re: What is a VMS Cluster  Re: What is a VMS Cluster   F ----------------------------------------------------------------------  # Date: Thu, 17 Apr 2003 22:09:23 GMT ) From: "lorabanks" <qxoupkbs@pvntokbr.com> . Subject: *&& << 1GB >> Flash Memory Drive (&(*@ Message-ID: <7464721cae6e8138ccdb2000e16d6817@news.teranews.com>  & Every units purchased... FREE SHIPPING   ++ 2GB / 1GB --> $699 / $334  3 >>>((((*** http://www.ebestshoppers.com *** )))<<<<    ------------------------------    Date: 17 Apr 2003 15:46:57 -0500; From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) $ Subject: Re: 7.1-1 H1 to 7.3: issue?3 Message-ID: <r$20t2URw8a3@eisner.encompasserve.org>   h In article <3e9edd9b$0$28763$626a54ce@news.free.fr>, PRSTSC::DTL <Didier.Morandi.nospam@Free.fr> writes: > Bob Koehler wrote:@ >>    Can they define "real time"?  I've known about 4 different >>    definitions. >   > F77 doing robots driving (DNC)      Sounds real to me.   R > (PS: I do not see either why there should be a problem, I even think that there B > is no need for compil/link, but it was worth asking, wasn't it?)  .    We did compile/link for two major upgrades:  F    VMS 2 didn't supply Fortran include files for all system constants,G    so a previous contractor had rolled their own.  We changed theirs to B    pick up the copy supplied with VMS 3 and wanted to make sure we@    didn't miss anything, also a few IO symbols actually changed.  D    VMS 4 was heavy enough to run somewhat slower on our 11/780's andB    we wanted to take advantage of the new Fortran compier's global5    optimizer, which more than made up the difference.    ------------------------------    Date: 17 Apr 2003 16:16:49 -0500+ From: kuhrt@encompasserve.org (Marty Kuhrt)  Subject: Average Uptimes by OS3 Message-ID: <AGK4g$sHIxee@eisner.encompasserve.org>   > For an interesting look at average uptimes by OS check out...   ( http://uptimes.wonko.com/stats.php?op=os  ; With the exception of one AIX machine, OpenVMSCluster leads < the pack with over 180+ days.  Sun clocks in at 39 days and < Windows averages about 3 days.  Plain old OpenVMS ranks 3rd.   ------------------------------  # Date: Thu, 17 Apr 2003 17:32:35 CDT / From: Charles Sandmann <sandmann@clio.rice.edu> " Subject: Re: Average Uptimes by OS- Message-ID: <3e9f2b83.sandmann@clio.rice.edu>   ) > With the exception of one AIX machine,    . What's it worth to you to make it go away? :-)  < Before they lost their database, I had a VAXStation 4000/M909 on VMS6.2 (DVAX1) which had been up close to 900 days ...   8 Alas, a 4 hour power outage drained the UPS a month ago.   ------------------------------    Date: 17 Apr 2003 19:22:58 -0500+ From: kuhrt@encompasserve.org (Marty Kuhrt) " Subject: Re: Average Uptimes by OS3 Message-ID: <ojeH+fRkgp2b@eisner.encompasserve.org>   _ In article <3e9f2b83.sandmann@clio.rice.edu>, Charles Sandmann <sandmann@clio.rice.edu> writes: * >> With the exception of one AIX machine,  > 0 > What's it worth to you to make it go away? :-) > > > Before they lost their database, I had a VAXStation 4000/M90; > on VMS6.2 (DVAX1) which had been up close to 900 days ...  > : > Alas, a 4 hour power outage drained the UPS a month ago.  : If I knew where the AIX was located I'd email them a plea. Ah, then again, probably not.     ; Since it is _only_ one machine, I don't think it's average  < should hold as much weight as the 12 systems averaged in the; OpenVMSCluster stats (of course, _I'm_ not biased at all!).   = I had a 4100 that was on the stats for a while, too.  It had  = over 500 days of uptime (it was the last remaining member of  < a four node cluster) until I tried to clear an error counter> and crashed the machine.  Ooops.  Not the preferred method for> clearing an error counter.  My Dublin cluster has been up for = 783 days, so I don't feel _so_ bad about pooching the 450 day  stat.   ? More VMS folks should get in on this (unless you system is like = my desktop at home and is powered up and shutdown every day). > This site is by no means an accurate representation of uptimesA across the industry, but since when does factual accuracy matter? B (digs at AH, specifically, and any Micro$lop product, in general,  purely intentional)  :^)   ------------------------------  # Date: Thu, 17 Apr 2003 20:41:48 CDT / From: Charles Sandmann <sandmann@clio.rice.edu> " Subject: Re: Average Uptimes by OS- Message-ID: <3e9f57dc.sandmann@clio.rice.edu>   < > If I knew where the AIX was located I'd email them a plea.! > Ah, then again, probably not.     A The AIX box is clio (see email above).  The reason it's a set of  : one is the default client doesn't build properly from the , distribution for AIX - so I had to fix it.    = > Since it is _only_ one machine, I don't think it's average  > > should hold as much weight as the 12 systems averaged in the= > OpenVMSCluster stats (of course, _I'm_ not biased at all!).   ? The only reason my VMS cluster ever goes down is extended power ; outages (thank you Reliant).  At this point it's not worth  = putting the cluster on wonko unless I can keep reliable power  to the building.   ------------------------------  # Date: Fri, 18 Apr 2003 02:03:58 GMT 4 From: brad@.gateway.2wire.net (Bradford J. Hamilton)" Subject: Re: Average Uptimes by OS? Message-ID: <i2Jna.529631$sf5.842635@rwcrnsc52.ops.asp.att.net>   a In article <ojeH+fRkgp2b@eisner.encompasserve.org>, kuhrt@encompasserve.org (Marty Kuhrt) writes:  <snip>@ >More VMS folks should get in on this (unless you system is like> >my desktop at home and is powered up and shutdown every day).? >This site is by no means an accurate representation of uptimes B >across the industry, but since when does factual accuracy matter?C >(digs at AH, specifically, and any Micro$lop product, in general,   >purely intentional)  :^)   N I just joined this evening, and my uptime did not significantly affect the VMSC average (something over 69 days - my machine has 69 days, as well).   A _________________________________________________________________ 0 Bradford J. Hamilton			"All opinions are my own"/ bMradAhamiPltSon@atMtAbi.cPoSm		"Lose the MAPS"    ------------------------------  % Date: Thu, 17 Apr 2003 22:31:14 -0700 % From: Dean Woodward <deanw@rdrop.com> " Subject: Re: Average Uptimes by OS( Message-ID: <3E9F8DA2.1040509@rdrop.com>   Bradford J. Hamilton wrote: P > I just joined this evening, and my uptime did not significantly affect the VMSE > average (something over 69 days - my machine has 69 days, as well).   F I've got access to two systems over 200 days; the system owners don't 2 allow me to run things like the uptime client. :-(  F I think the uptime stats are a lot more impressive when you factor in H the CPU load- multiplying average uptime by average CPU load puts VMS / / VMS cluster at or very near the top, I think...    ------------------------------  + Date: Thu, 17 Apr 2003 20:03:45 +0000 (UTC) 3 From: "Richard Maher" <maher_rj@hotspamnotmail.com> C Subject: Cache Fusion? (Was: Re: Excellent new OpenVMS Testimonial) / Message-ID: <b7n1b0$54o$1@titan.btinternet.com>    Hi,   4 Great testimonial. Thanks for posting the reference.  K One quick question though. The article says they're running Oracle9i RAC on K clustered Alphas; does anyone know if Fraport has implemented Cache Fusion? J (Has *anyone* implemented Cache Fusion on VMS?) I'd love to hear about anyH experiences with this amazing functionality on VMS. I still have trouble- getting my head around how it can make sense.   F Does anyone know if Fraport also have OracleNi or SQL Server databases runing on a Windows platform?    Regards Richard Maher.  H PS. It's great to see another *cluster* success story and that, at least4 part of, Oracle is playing to a cluster's strengths!  < Sue Skonetski <susan_skonetski@hotmail.com> wrote in message6 news:857e9e41.0304141436.382fdd0@posting.google.com... > Dear Newsgroup,  > C > Warren has put up a new customer testimonial on the web site. The C > customer is  Fraport (Frankfurt, Germany airport).  Please take a G > moment to look at this.  I think it is excellent. Marc Courchesne and 8 > Guenter Krieble are the HP folks responsible for this. >  > Warm Regards,  > Sue    ------------------------------  % Date: Thu, 17 Apr 2003 11:44:49 -0700 # From: "Tom Linden" <tom@kednos.com> B Subject: RE: DCL question (of the day)- null byte in string symbol9 Message-ID: <CIEJLCMNHNNDLLOOGNJIGEEMHAAA.tom@kednos.com>    >-----Original Message----- A >From: briggs@encompasserve.org [mailto:briggs@encompasserve.org] ( >Sent: Thursday, April 17, 2003 10:17 AM >To: Info-VAX@Mvb.Saic.ComC >Subject: RE: DCL question (of the day)- null byte in string symbol  >  > ? >In article <CIEJLCMNHNNDLLOOGNJIKEEIHAAA.tom@kednos.com>, "Tom ! >Linden" <tom@kednos.com> writes: L >> That doesn't make sense.  Why would parsing continue after it encountered >> a "!"  ?  > F >Because parsing is not a monolithic process.  You do apostrophe-basedD >substitution first because the result of an apostrophe substitutionD >could contain an exclamation point that would mark the remainder of >the command as a comment.  B Well, I don't know how they did it, but I had assumed that once an apostrophe were L detected, then "!" would have had a different significance until the closingB apostrophe were detected.  At least that his I would have done it.  G In your example below, when parsing the second line, q has already been 
 subsitutedJ so on the scan you know you need to look for a double quote so the ! would not A be treated  as a comment initiator, unless the parsing is ad hoc.    > @ >Since you're going to look for exclamation points later anyway,? >you don't look for exclamation points while you're looking for + >apostrophes.  It would be a waste of time.  > E >Since apostrophe substitution could produce a double quotation mark, B >thus turning an apparent comment into a part of a string literal,E >you can't legitimately pre-process comment markers anyway.  Not only 4 >would it be a waste of time, it would be incorrect. >  >$	q = """"  >$	b = 'q'as!df'q'
 >$	show sym b  > 
 >	John Briggs  >  >---' >Incoming mail is certified Virus Free. ; >Checked by AVG anti-virus system (http://www.grisoft.com). A >Version: 6.0.471 / Virus Database: 269 - Release Date: 4/10/2003  >  --- & Outgoing mail is certified Virus Free.: Checked by AVG anti-virus system (http://www.grisoft.com).@ Version: 6.0.471 / Virus Database: 269 - Release Date: 4/10/2003   ------------------------------  % Date: Thu, 17 Apr 2003 15:46:51 -0400  From: norm.raphael@metso.comB Subject: Re: DCL question (of the day)- null byte in string symbol? Message-ID: <OFD4FB0E32.A4C8914C-ON85256D0B.006B7764@metso.com>   ? As to these, I have my answer and am content with the behavior. F The null is documented as to what it will do and why it should not (or could) be intentionally F introduced.  It terminates parsing, not inputting or verification of a line. % I ask for no change in that behavior.   K In my case, it was introduced into a variable by a read operation.  The fix  for the problem 3 was to understand my input and program accordingly.   ; Thanks, Richard, for pointing me back to the documentation. C IMO there is not even a documentation error involved, although some  discussion, perhaps K in the DCL dictionary (if it is not already there - I'm not taking the time  to check) might help, as1 most of us benefit immeasurably by good examples.    -Norm   B From:  "Richard Brodie" <R.Brodie@rl.ac.uk> on 04/17/2003 01:05 PM  6 Please respond to "Richard Brodie" <R.Brodie@rl.ac.uk>   To:    Info-VAX@Mvb.Saic.Com cc:   E Subject:    Re: DCL question (of the day)- null byte in string symbol       K <mckinneyj@cpva.saic.com> wrote in message news:Hb$PRPjcfP0i@cpva.saic.com.  . 5 > In article <KXwKUXMAhyLh@eisner.encompasserve.org>,   J > could be a bug, could be a feature... don't know. But it appears to have/ > behaved this way since early versions of VMS.    I'll go for feature:  M http://groups.google.com/groups?selm=8808031313.AA27433%40ucbvax.berkeley.edu   G http://h71000.www7.hp.com/doc/731FINAL/6489/6489pro_003.html#rules_sect      [=]+  5 From: briggs@encompasserve.org on 04/17/2003 01:32 PM   * Please respond to briggs@encompasserve.org   To:    Info-VAX@Mvb.Saic.Com cc:   E Subject:    Re: DCL question (of the day)- null byte in string symbol     < In article <b7mmsk$spq@newton.cc.rl.ac.uk>, "Richard Brodie" <R.Brodie@rl.ac.uk> writes:  > , > <mckinneyj@cpva.saic.com> wrote in message" news:Hb$PRPjcfP0i@cpva.saic.com...6 >> In article <KXwKUXMAhyLh@eisner.encompasserve.org>, > K >> could be a bug, could be a feature... don't know. But it appears to have 0 >> behaved this way since early versions of VMS. >  > I'll go for feature: >  > M http://groups.google.com/groups?selm=8808031313.AA27433%40ucbvax.berkeley.edu   I > http://h71000.www7.hp.com/doc/731FINAL/6489/6489pro_003.html#rules_sect   * The relevant part of the latter URL being:  D   "You cannot specify null characters (<NUL>) on a DCL command line,>    even if you enclose the null character in quotation marks."  F That means it's neither a bug nor a feature.  It's out in the no man's? land of undefined behavior.  You are not allowed to write a DCL 3 program for which the actual behavior would matter.   F It seems clear that cases in which a <NUL> is subtly embedded in a DCL< command line through the use of symbol substitution are also@ intended to be forbidden.  Any "bug" here is thus reduced to the: level of a documentation error rather than of a DCL error.                John Briggs   ------------------------------    Date: 17 Apr 2003 15:57:08 -0500; From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) B Subject: RE: DCL question (of the day)- null byte in string symbol3 Message-ID: <8GDvUQbMk370@eisner.encompasserve.org>   _ In article <CIEJLCMNHNNDLLOOGNJIKEEIHAAA.tom@kednos.com>, "Tom Linden" <tom@kednos.com> writes:  > K > That doesn't make sense.  Why would parsing continue after it encountered 
 > a "!"  ?      &    Parsing is done first to find the !   ------------------------------  # Date: Thu, 17 Apr 2003 18:04:41 GMT % From: "BCbuyer" <BCbuyer@hotmail.com> N Subject: Help! Bizzare behaviour of C RTL function stat - issue for SAMBA port7 Message-ID: <Z0Cna.596$ja.111482@news2.calgary.shaw.ca>   L There has been much discussion in the SAMBA-VMS mailing list about improvingJ performance for directory browsing. Much of the attention has been focusedL on the cost of the stat() function from the DEC C RTL. Work has been carriedH out to cache file information such as size, creation date, etc. This hasJ lead to remarkable performance improvements, except for that first read of the directory.  I The stat() function does a lookup of a given file by name to retrieve all F the relevant information. "Knowing" that this was a problem, I wrote aK simple program to mimic the behaviour of the SAMBA code. It uses opendir(), B then loops over the directory with readdir() to retrieve a list ofJ filenames, then closedir(). A separate loop invokes stat() to retrieve theJ information for each file in the list generated earlier. This organization> is dictated by general SAMBA design - we have no problem here.  L I wrote a simple program that executes just these C function calls against aH directory specified with Unix syntax. One of the effects of this is thatH only the top-level files are seen - without version numbers at that. TheC program generated approximately three I/O per top-level file in the F directory. I expected this given the overhead of file name resolution.  I My idea was to use RMS calls to return the file name, FID, DID and volume L name in the first loop and then use an RMS open by NAM block with the FID toL bypass most of the overhead. I wrote another simple program to test this outL and saw approximately one I/O per top-level file in the directory. Again, as	 expected.   F Going back to the actual SAMBA server program, I ran a careful test toK verify the I/O load. To my surprise, a SHOW PROCESS /CONTINUOUS showed that E it was generating approximately one I/O per top-level file during the ! servicing of a directory request!   L I don't understand how it is possible for a loop of stat() calls to generateL so few I/O. Further, I don't understand why this behaviour is not consistentH with the simplified program. All tests were run against a directory withG over 800 top level files. The .DIR for this directory was less than 128  blocks in size.     L Does anybody know for fact how the stat() function is implemented on currentL VMS systems? What would affect the number of I/Os it costs for a given call.K If it makes a difference, we are talking about an Alpha and code built with K the current DEC C compiler and run with all the latest, greatest patches on  version 7.2-1.     Thanks,   Peter   ------------------------------   Date: 17 Apr 2003 14:12 CDT ' From: carl@gerg.tamu.edu (Carl Perkins) R Subject: Re: Help! Bizzare behaviour of C RTL function stat - issue for SAMBA port- Message-ID: <17APR200314125147@gerg.tamu.edu>   ) "BCbuyer" <BCbuyer@hotmail.com> writes... M }There has been much discussion in the SAMBA-VMS mailing list about improving K }performance for directory browsing. Much of the attention has been focused M }on the cost of the stat() function from the DEC C RTL. Work has been carried I }out to cache file information such as size, creation date, etc. This has K }lead to remarkable performance improvements, except for that first read of  }the directory.   / Yes, it is a LOT faster. (Thank you Jean-Yves!)   J }The stat() function does a lookup of a given file by name to retrieve allG }the relevant information. "Knowing" that this was a problem, I wrote a L }simple program to mimic the behaviour of the SAMBA code. It uses opendir(),C }then loops over the directory with readdir() to retrieve a list of K }filenames, then closedir(). A separate loop invokes stat() to retrieve the K }information for each file in the list generated earlier. This organization ? }is dictated by general SAMBA design - we have no problem here.  } M }I wrote a simple program that executes just these C function calls against a I }directory specified with Unix syntax. One of the effects of this is that I }only the top-level files are seen - without version numbers at that. The D }program generated approximately three I/O per top-level file in theG }directory. I expected this given the overhead of file name resolution.  } J }My idea was to use RMS calls to return the file name, FID, DID and volumeM }name in the first loop and then use an RMS open by NAM block with the FID to M }bypass most of the overhead. I wrote another simple program to test this out M }and saw approximately one I/O per top-level file in the directory. Again, as 
 }expected. } G }Going back to the actual SAMBA server program, I ran a careful test to L }verify the I/O load. To my surprise, a SHOW PROCESS /CONTINUOUS showed thatF }it was generating approximately one I/O per top-level file during the" }servicing of a directory request!  B As you noted above, the current version of SAMBA does caching. Are$ you certain you weren't seeing this?  M }I don't understand how it is possible for a loop of stat() calls to generate M }so few I/O. Further, I don't understand why this behaviour is not consistent I }with the simplified program. All tests were run against a directory with H }over 800 top level files. The .DIR for this directory was less than 128 }blocks in size. } M }Does anybody know for fact how the stat() function is implemented on current M }VMS systems? What would affect the number of I/Os it costs for a given call. L }If it makes a difference, we are talking about an Alpha and code built withL }the current DEC C compiler and run with all the latest, greatest patches on }version 7.2-1.  }  } Peter   J The stat() function probably works through RMS. RMS caches a lot of stuff.E It is possible that for the directory you were using, the RMS caching K eliminated a lot of the disk I/O. For example, when it reads the INDEXF.SYS H file it may not be reading just one block. If it reads in several blocksG and caches them for later accesses then it could reduce the I/O count a I lot if the files in your directory have sequential, or nearly sequential, 	 file IDs.   K I have tested the case of using an opendir(), loop on readdir(), closedir() J and then sorting all the files by FID (readdir() returns both the filenameF and the FID stored in the .DIR file). Then directly opening INDEXF.SYSJ and using QIOs to read the file headers (IO$_READVBLK to directly read theG block of the file in which that header is located) to get the info that H the stat() function returns. This is very fast compared to using stat(),L particularly when using asynch QIOs with two outstanding at a time. It worksN quite well, using ony about 16% of the CPU time of a stat() in a lib$find_fileL loop and taking slighty under 50% of the elapsed time for my test directory.O Although my code is specific to ODS-2, it should need only minor modifications, L if any, for ODS-5. It would be faster in some cases if it read more than oneM block per QIO and cached them until a new read was needed (this would be made J somewhat easier since I sort the FIDs - if you hit one past the end of theL set you read in, then you know that you won't be doing a read to a block youJ already read in before). I didn't record I/O counts for my tests, just CPU+ time, so I'm not certain how that compares.    --- Carl   ------------------------------  # Date: Thu, 17 Apr 2003 19:51:52 GMT % From: "BCbuyer" <BCbuyer@hotmail.com> R Subject: Re: Help! Bizzare behaviour of C RTL function stat - issue for SAMBA port= Message-ID: <sBDna.744846$Yo4.71736026@news1.calgary.shaw.ca>   4 "Carl Perkins" <carl@gerg.tamu.edu> wrote in message' news:17APR200314125147@gerg.tamu.edu... + > "BCbuyer" <BCbuyer@hotmail.com> writes... E > }There has been much discussion in the SAMBA-VMS mailing list about 	 improving E > }performance for directory browsing. Much of the attention has been  focused G > }on the cost of the stat() function from the DEC C RTL. Work has been  carried K > }out to cache file information such as size, creation date, etc. This has J > }lead to remarkable performance improvements, except for that first read of > }the directory.  > 1 > Yes, it is a LOT faster. (Thank you Jean-Yves!)    Add my thanks to that.     (snip)  I > }Going back to the actual SAMBA server program, I ran a careful test to I > }verify the I/O load. To my surprise, a SHOW PROCESS /CONTINUOUS showed  thatH > }it was generating approximately one I/O per top-level file during the$ > }servicing of a directory request! > D > As you noted above, the current version of SAMBA does caching. Are& > you certain you weren't seeing this?  I Positive. I ran the test with and without the debugger. I started the I/O H counting just as the command line client issued the DIR command. I could@ also see that this is the code stuffing the data into the cache.   > L > The stat() function probably works through RMS. RMS caches a lot of stuff.G > It is possible that for the directory you were using, the RMS caching B > eliminated a lot of the disk I/O. For example, when it reads the
 INDEXF.SYSJ > file it may not be reading just one block. If it reads in several blocksI > and caches them for later accesses then it could reduce the I/O count a K > lot if the files in your directory have sequential, or nearly sequential,  > file IDs.  >   J That would be fine, except my test program, which was implemented with theI same C RTL calls, generated the higher I/O counts one would expect. Also, H the files in question were created over a space of years - more than 800 non-sequential FIDs.  L This is the big problem I have. Even when I come up with a mechanism wherebyL stat() can work faster than expected, I cannot explaion why the test program is so slow.     B > I have tested the case of using an opendir(), loop on readdir(),
 closedir()L > and then sorting all the files by FID (readdir() returns both the filenameH > and the FID stored in the .DIR file). Then directly opening INDEXF.SYSL > and using QIOs to read the file headers (IO$_READVBLK to directly read theI > block of the file in which that header is located) to get the info that J > the stat() function returns. This is very fast compared to using stat(),H > particularly when using asynch QIOs with two outstanding at a time. It works B > quite well, using ony about 16% of the CPU time of a stat() in a
 lib$find_file C > loop and taking slighty under 50% of the elapsed time for my test 
 directory.B > Although my code is specific to ODS-2, it should need only minor modifications,J > if any, for ODS-5. It would be faster in some cases if it read more than one J > block per QIO and cached them until a new read was needed (this would be madeL > somewhat easier since I sort the FIDs - if you hit one past the end of theJ > set you read in, then you know that you won't be doing a read to a block you L > already read in before). I didn't record I/O counts for my tests, just CPU- > time, so I'm not certain how that compares.  >   J It was your work that turned me onto this! I also did some I/O comparisonsJ of stat() vs RMS vs Files-11 calls. The RMS calls gernerated significantlyL fewer I/Os than the Files-11 calls. Both of these mechanisms generated but a  fraction of the I/Os stat() did.     Cheers,   Peter Smode   ------------------------------  % Date: Thu, 17 Apr 2003 13:46:05 -0500 / From: "Stuart, Ed" <Ed.Stuart@austinenergy.com> ? Subject: How to access DB2 on an AS/400 from Oracle on OpenVMS? T Message-ID: <92EFB80E551BD511B39500D0B7B0CDCC0CEF970E@ohms.electric.ci.austin.tx.us>  J This message is in MIME format. Since your mail reader does not understand< this format, some or all of this message may not be legible.  ' ------_=_NextPart_001_01C30511.99F94920  Content-Type: text/plain  
 Hello all;  J We are looking for a driver that will allow us to access a DB2 database onF an AS/400 from an Oracle database (8.1.7) on OpenVMS (7.2-1). From theG Windows world we use IBM Client Access, but I haven't been able to find ( Client Access for OpenVMS.  Suggestions?   EdE **Please apply a generous amount of all the usual disclaimers here.**     ' ------_=_NextPart_001_01C30511.99F94920  Content-Type: text/html + Content-Transfer-Encoding: quoted-printable   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">  <HTML> <HEAD>9 <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =  charset=3Dus-ascii">@ <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
 5.5.2653.12"> E <TITLE>How to access DB2 on an AS/400 from Oracle on OpenVMS?</TITLE>  </HEAD>  <BODY>  2 <P><FONT SIZE=3D2 FACE=3D"Arial">Hello all;</FONT> </P>  H <P><FONT SIZE=3D2 FACE=3D"Arial">We are looking for a driver that will =H allow us to access a DB2 database on an AS/400 from an Oracle database =F (8.1.7) on OpenVMS (7.2-1). From the Windows world we use IBM Client =; Access, but I haven't been able to find Client Access for = & OpenVMS.&nbsp; Suggestions?</FONT></P>  * <P><FONT SIZE=3D2 FACE=3D"Arial">Ed</FONT>G <BR><FONT SIZE=3D2 FACE=3D"Arial">**Please apply a generous amount of = ( all the usual disclaimers here.**</FONT> </P>   </BODY>  </HTML> ) ------_=_NextPart_001_01C30511.99F94920--    ------------------------------  % Date: Thu, 17 Apr 2003 23:23:14 -0400 ' From: Stuart Fuller <stufuller@usa.net>   Subject: Re: HP Shipping Charges0 Message-ID: <33rn7b.rat.ln@dadsys2.fuller.local>   Chris Scheers wrote:  E > This information may already be widely known, but it was a bit of a  > surprise to me.  > J > In the past (Ah, the good old days.  <grin>), whenever I placed an orderJ > with DEC for parts, the shipping, aka "Transportation Charge" was always) > reasonable, usually just a few dollars.  > I > I recently placed an order with HP for a $33 part, and was invoiced for  > a $25 transportation charge. > G > When I asked about this, I was told "You know, if you had paid with a 2 > credit card, the shipping would have been free." > H > I responded that, when placing the order, I had explicitly asked if itF > made any difference how I paid, and was told that it didn't make any
 > difference.  > I > The response to this was: "Oh, we don't tell the customers about that."  > I > ----------------------------------------------------------------------- & > Chris Scheers, Applied Synergy, Inc. > D > Voice: 817-237-3360            Internet: chris@applied-synergy.com >   Fax: 817-237-3074   I Was the $25 charge mentioned at the time of the order?  If not, then I'd   not pay it.    --             Stu    ------------------------------    Date: 17 Apr 2003 14:31:06 -0400% From: Thien-Thi Nguyen <ttn@glug.org> # Subject: Re: HTML favourite editor? ( Message-ID: <7gel41x9g5.fsf@gnufans.net>  % "Tom Linden" <tom@kednos.com> writes:   7 > Now that is good news!  Is based on what Roar Throns  > [roart@nvg.ntnu.no]   ? in part.  the bulk of the work is adapting (to) emacs' somewhat B idiosyncratic build process, which relies on the C preprocessor atH configure-time to not insert whitespace between tokens, as is apparentlyG (now) allowed by the C standard(s); and to do all this in a way that is > easy to fold into the GNU cvs repo, proper, over the long run.  A the rest of the effort is battling my huge ignorance of vms.  for H example, after having solved the above problem by bundling an extractionG of gcc's "traditional cpp" (see http://www.glug.org/alt/tradcpp.zip), i F learned that there are switches to the system C compiler that probablyC can achieve the same result.  argh!  still, i feel lucky that i can E build on the work of Throns and Levitte, learning interesting things  in the process.    anyway, back to lurking...   thi    ------------------------------  % Date: Thu, 17 Apr 2003 17:10:31 -0400 0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>% Subject: Interesting VMS ad in Google / Message-ID: <3E9F183B.B1388697@vl.videotron.ca>   L Did a search for some posts since my current ISP loses a significant portion of NNTP traffic.  J When i popped the thread where I asked a qestion to see what the responses) were, here are the sponsored links I got:g   Sponsored Links : F Qt to Motif Migration  Preserve your investment in Motif Migrate your  applications to Qt  www.ics.comN                                                                                       yD OpenVMS System Management  Have your system managed by HP Certified professionals.  www.stanq.com  K X-Designer  Advanced cross platform GUI Builder Download a free evaluationR copy  www.ist-inc.com   -----A    L I find it most interesting that it would have known about my interest in VMSK since (the machine used  to retrieve the messages on google is a mac) and I P only made a mention of "VMS" once in the question (which was X-windows related).  L It seems like Google has added lots of smarts to its software. I wonder whatK sort of arrangement www.stanq.com has with Google in terms of deciding whatn  triggers their "sponsored link".   ------------------------------  % Date: Thu, 17 Apr 2003 13:04:56 -0500a/ From: "Stuart, Ed" <Ed.Stuart@austinenergy.com>f3 Subject: RE: Mirroring versus Shadowing - any diff?iT Message-ID: <92EFB80E551BD511B39500D0B7B0CDCC0CEF970A@ohms.electric.ci.austin.tx.us>  I I thought that mirroring was used when the reading/writing was being doneiG from one disk and the transactions were applied to the "mirror" disk bymI whatever was handling the mirroring, and that shadowing was used when ther3 read/write operations were spread across the disks.    EdE **Please apply a generous amount of all the usual disclaimers here.**r   ------------------------------    Date: 17 Apr 2003 12:19:26 -0700# From: wmarsh@etrade.com (Wil Marsh) 3 Subject: Re: Mirroring versus Shadowing - any diff?i< Message-ID: <36f52bd6.0304171119.fc53438@posting.google.com>  g JF Mezei <jfmezei.spamnot@vl.videotron.ca> wrote in message news:<3E9E76FA.429A316C@vl.videotron.ca>...S > Nic Clews wrote:F > > Which is better for you, depends. Offhand I don't know of any diskD > > technology books that discuss RAID levels in any detail, anyone? > K > Host based shadowing has advantages in trems of cluster sanity because it M > simplifies the process of keepin data integrity incase of a node's failure.  > P > Hardware based mirroring/shadowing offloads some processing from the operatingN > system. It also isolates the OS from the intricacies of the mirroring, whichN > may be problematic when a problem arises if the disk array hides the problem > from the OS. > N > Host based shadowing does have a neat capability though: if you have 4 nodesO > in a cluster, with one node doing a hell of a lot of "reads", it can actuallynH > use MSCP to spread the reads across all nodes giving you built-in loadO > balancing. If you have 4 drives in an array, each node then has a single path H > to the array so there is only so many concurrent IOs that can be done. > N > Another aspect of VMS host based shadowing is that when your write completesM > at the application level, you are garanteed that the write completed on all.K > drives. That si not always the case on a disk array, especially if it has:N > fancy caching/buffering. (I guess that problem can be alleviated with a goodO > UPS system for the disk array to ensure that any data in the buffer does makeI
 > it to disk.h  F shadowing has an advantage if there are multiple controllers involved.D by putting shadow set members on different controllers you avoid theD issue of controller failure causing an outage, where mirror sets areA on the same controller. also shadow members can be dismounted fora  backup while mirror sets cannot.   ------------------------------  # Date: Thu, 17 Apr 2003 19:34:02 GMT % From: "BCbuyer" <BCbuyer@hotmail.com> 3 Subject: Re: Mirroring versus Shadowing - any diff?o< Message-ID: <KkDna.147530$vs.16886620@news3.calgary.shaw.ca>  @ First question is exactly what controllers are you going to makeI dual-redundant. If they are CI controllers, are they HSJ40, HSJ50, HSJ80? G Are you using fibre based HSG80s? Or are you using some other flavor ofKF interconnect completely? This will have great impact on the perfomance# characteristics of what you can do.s  G Certain things apply regardless of controller. For example, rebooting amH node, either standalone or in a cluster, will cause VMS operating systemG software shadow sets to fall out of steady state. You'll either have todK suffer through a shadow merge or shadow copy. If you have a controller that-L supports logging (e.g. HSJ50s with the right firmware), the operating systemL will often be able to engage in a mini-merge, theoretically lowering the I/OF load of a full merge. This will only work if the two volumes have beenK created on the same controller pair. However, you are going to be hammeringeJ your disks and cluster adapters while all this goes on. If you are runningK many volumes in this sort of configuration, it can have devastating impactstJ on performance. Further, during this time frame, you have no protection inB case of failure of the volume that is the source of the shadow setG resolution. This is also the dirty little secret of multi-site clustersn* relying exclusively on software shadowing.  I Obviously, software shadowing has certain things going for it, especiallyo> the relative certainty of I/O completion. However, it is not a one-size-fits-all solution.t  J Controller-based hardware mirroring has great performance characteristics;F however, I would never count on it unless I was running dual-redundantA controllers with battery backup for the cache. One advantage thatiJ controller-based mirroring has is that if a mirror setneeds to be rebuilt,L all the I/O stays on the SCSI buses of the controllers and never goes out onK the storage/cluster interconnect back to the VMS host (and around the otherlE controllers). With this, you do not place load on the cluster/storage B adapter of the VMS host nor do you clutter up the bandwidth of theJ interconnect, effectting all hosts and controllers in the installation. IfG you are used to software shadow set resolution, you'll be amazed at how = little time these operations take - minutes instead of hours.s  G To me, the more significant advantage is that the state of VMS hosts isoI irrelavent to the steady state characteristics of the mirror set. You candL shut down your whole cluster and it doesn't matter since the controllers areG always running. Even normal shutdown of dual-redundant controllers wontrD affect the mirrror set. Personnaly, I have never seen a failure of aL controller-based mirror set without a failure of one of the disk drives that implements it.  K Another nice thing about controller-based mirror sets is automatic recoveryUJ of disk failure. When setting things up, you should place a few extra diskK drives in the shelves and tell the controller that these are members of thenL SPARESET and configure the mirror set to use this (AUTOSPARE, I think). If aH drive in a mirror set fails in this setup, the controller will check theD SPARESET to see if an identical drive is present. If there is, it isK automatically brought into the mirror set and a copy operation is initiatedeK on the controller. The bad disk is always put into a separate collection of K disks. These activities are logged out by the controller on the serial linetH and back to VMS hosts with an SCS message. The activity is logged in theJ error log of the VMS system. If you have DECEVENT and call logging enabledJ though DSNlink, it will automatically log a call to field service. You can@ also tell DECEVENT to send you a page or an e-mail. This sort ofG functionality is not available with software shadowsets without alot of  custom work.  L With SAN controllers like the HSG80 and appropriate software (i.e. DRM), youA can mirror volumes across multiple sites without operating systemtL intervention. However, this is a dangerous thing to do in isolation with VMSG systems in a milti-site cluster. In this case, the best balance of highhK performance and reliability comes with combining controller-based mirroringfH across multiple sites with VMS software volume shadowing. To really makeI this sing you need v7.3 features and some careful planning. Since you aretJ only now getting into dual-redundant controllers, I'll assume that you are9 not running with this sort of technology or requirements._  H Certain situations may call for a combination of shadow sets with one orH more members being controller-based mirror sets. If you are one of thoseL folks that splits shadow sets for certain operations, simply make one of theK memebers a controller-based mirror set and always leave it as the surviving> member.o  J With controller-based mirrors, make sure that you put each of its memebersK on a different shelf. This helps with performance mostly, but also is a bitsH of a reliability issue. Also, don't put all your SPARESET members on theI same shelf. Rememeber to that each controller pair needs its own SPARESET  colleciton.p     Cheers,e  Peter Smode   ------------------------------  % Date: Thu, 17 Apr 2003 19:42:08 -0000v! From: Z  <zarlenga@conan.ids.net>s3 Subject: Re: Mirroring versus Shadowing - any diff?-/ Message-ID: <v9u0sg1fdf999b@corp.supernews.com>:  ! Z <zarlenga@conan.ids.net> wrote:h; : My new employer's hardware expert says shadowing is host-a= : based while mirroring is HSC-based.  (I will soon be fixing$   Correction, HSZs are used.  . Thank you to all who replied with information.   -- e  *     "We Gave Peace a Chance; We Got 9/11."   ------------------------------  % Date: Thu, 17 Apr 2003 17:21:00 -0400i0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>3 Subject: Re: Mirroring versus Shadowing - any diff?e/ Message-ID: <3E9F1AB0.774415B7@vl.videotron.ca>i   Wil Marsh wrote:H > shadowing has an advantage if there are multiple controllers involved.F > by putting shadow set members on different controllers you avoid the0 > issue of controller failure causing an outage,  L Is this a temporary issue though ? Won't SANs eventually have the ability toN have redundant controllers with automatic failover ? (But right now, I realiseG this is an advantage since those that do have dual controllers, seem toiH require manual reconfig to make the failover happen. (is this correct ?)   > where mirror sets areiC > on the same controller. also shadow members can be dismounted fore" > backup while mirror sets cannot.  M Conceptually, would it be possible to have the SAN take one of the drives outsM and present a new additional drive to VMS so that VMS could use it to perform- the backup ?  I VMS has the ability to see new drives appear dynamically (when a new nodejK boots into the cluster, making its drives available for instance). So a SANCG should be able to make a new drive appear dynamically to VMS, correct ?-  I What about when you want to widthdraw that drive and place it back in the:L mirror, can VMS totally forget about a drive, or does it always remain known' (marked offline/noavailable/whatever) ?t  H Another question: in the above scenario, is it possible from DCL to sendI management commands to a SAN to execute that split of the  mirror set and   presentation of an extra drive ?  N (eg: to automate the backup, you don't want to have to have someone physicallyM telnet to the SAN to reconfigure the mirror array, then thell the operator to N mount the new drive, execute backup and call him again once backyup is done so4 that the drive can be re-integrated into the mirror.  L With VMS host based shadowing, that is just a simple mount/dismount command.   ------------------------------  % Date: Thu, 17 Apr 2003 23:04:08 +0100w9 From: Alan Adams <alan.adams@orchard-way.freeserve.co.uk>n3 Subject: Re: Mirroring versus Shadowing - any diff?h? Message-ID: <77cec3e44b.Alan.Adams@orchard-way.freeserve.co.uk>   . In message <3E9F1AB0.774415B7@vl.videotron.ca>;           JF Mezei <jfmezei.spamnot@vl.videotron.ca> wrote:    > Wil Marsh wrote:J > > shadowing has an advantage if there are multiple controllers involved.H > > by putting shadow set members on different controllers you avoid the2 > > issue of controller failure causing an outage, > N > Is this a temporary issue though ? Won't SANs eventually have the ability toP > have redundant controllers with automatic failover ? (But right now, I realiseI > this is an advantage since those that do have dual controllers, seem to=J > require manual reconfig to make the failover happen. (is this correct ?) >  > > where mirror sets are E > > on the same controller. also shadow members can be dismounted forV$ > > backup while mirror sets cannot. > O > Conceptually, would it be possible to have the SAN take one of the drives out O > and present a new additional drive to VMS so that VMS could use it to performb > the backup ? > K > VMS has the ability to see new drives appear dynamically (when a new node:M > boots into the cluster, making its drives available for instance). So a SANaI > should be able to make a new drive appear dynamically to VMS, correct ?. > K > What about when you want to widthdraw that drive and place it back in theaN > mirror, can VMS totally forget about a drive, or does it always remain known) > (marked offline/noavailable/whatever) ?e  3 As far as I know, it's there until the next reboot.i  J This could well re-open a discussion of a few months ago about disks whichJ exist but are not available. (In DCL, before mounting a disk, it's as well to check   1 does it exist? 2 is it availablen 3 is it mountedp. And give up on the first unfavourable answer.)   > J > Another question: in the above scenario, is it possible from DCL to sendK > management commands to a SAN to execute that split of the  mirror set and " > presentation of an extra drive ? > P > (eg: to automate the backup, you don't want to have to have someone physicallyO > telnet to the SAN to reconfigure the mirror array, then thell the operator todP > mount the new drive, execute backup and call him again once backyup is done so6 > that the drive can be re-integrated into the mirror. > N > With VMS host based shadowing, that is just a simple mount/dismount command.   -- X
 Alan Adams& alan.adams@orchard-way.freeserve.co.uk http://www.nckc.org.uk/e   ------------------------------  + Date: Thu, 17 Apr 2003 23:28:04 +0000 (UTC)f) From: Dan Foster <dsf@globalcrossing.net>o3 Subject: Re: Mirroring versus Shadowing - any diff?x3 Message-ID: <slrnb9ue44.t9g.dsf@gaia.roc2.gblx.net>N  a In article <3E9F1AB0.774415B7@vl.videotron.ca>, JF Mezei <jfmezei.spamnot@vl.videotron.ca> wrote:P > Wil Marsh wrote:I >> shadowing has an advantage if there are multiple controllers involved.cG >> by putting shadow set members on different controllers you avoid thea1 >> issue of controller failure causing an outage,o > N > Is this a temporary issue though ? Won't SANs eventually have the ability toP > have redundant controllers with automatic failover ? (But right now, I realiseI > this is an advantage since those that do have dual controllers, seem to J > require manual reconfig to make the failover happen. (is this correct ?)  K Well, unless I'm misunderstanding your comment, IBM's ESS ("Shark") already7G has that feature. For instance, the ESS has two internal host computers8B running special software and firmware with a large cache (4 GB perK controlling internal system) in a fully redundant mesh setup. All disks arenJ available in both internal (non-user visible) clusters; if you powered offH one cluster (one half of the ESS), you wouldn't notice a single thing on the served filesystems.a  I We know because we went through a cluster failure recently related to theaG physical move of the unit to a new datacenter. Zero downtime; the othernI cluster had some sort of cache coherency so it just took over the pendingdA writes, completed that, and continued as the new primary internalfK controller for all the active disks. Totally transparent. All this happenedsF internally to the unit; nothing visible at all to the servers that had$ volumes served from the ESS over FC.  F High price for that kind of redundancy and bulletproofness but it just" simply works, and works very well.  . http://www.storage.ibm.com/disk/ess/index.html  H Picture of our ESS (model F20) with the front doors temporarily open forG some additional pre-return to service inspection after a physical move.r  / http://home.globalcrossing.net/~dsf/our-ess.jpgo  G Top four drawers on both sides are the SSA160 disk drawers; more in thepK back on other side. Below...don't recall, maybe the batteries? I'd rememberBH if I was on-site. Below that are the two host systems; each one hosts an internal ESS cluster.n  E The blue cables are the SSA160 cables from the disk drawers. Near the J bottom are the orange FC-AL cables running from the various FC controllersH in the host bay adapter slots. Very bottom are the fans and probably the PSUs inside.  J Top right picture shows six thick wires running straight down... these areG the 30 amp electrical circuits, although I think only three are for thehJ Shark, and the others are for our nearby large systems. (You can't see theI large NEMA electrical connectors that leads to the large systems and ESS,tH but if you've seen a large server before, you'll know what they're like.6 Huge and one doesn't lick the capacitors in them :-) )  F In the middle is a panel with the emergency power kill switch, and theH normal power on/off switch. Below that are three groups of two LEDs. TwoJ groups of them are green and on -- that's good. The third group of LEDs onE the bottom are off -- that's the "error messages present" ones. Seems  healthy.  J Well designed unit; right below the disk drawers are a small black lookingE 'gap'... that's actually slide-out drawers. IBM CEs (hardware serviceiJ engineers) uses left to store CDs, floppies, and other small and importantH things. They use the right slide-out drawer to put their Thinkpad laptopF running a special service personnel-only application for the ESS while they're working on the unit.   -Dan   ------------------------------    Date: 17 Apr 2003 19:53:28 -07001 From: keithparris_NOSPAM@yahoo.com (Keith Parris)m3 Subject: Re: Mirroring versus Shadowing - any diff?n< Message-ID: <cf15391e.0304171853.aef73bd@posting.google.com>  i "BCbuyer" <BCbuyer@hotmail.com> wrote in message news:<KkDna.147530$vs.16886620@news3.calgary.shaw.ca>...rI > Certain things apply regardless of controller. For example, rebooting alJ > node, either standalone or in a cluster, will cause VMS operating systemI > software shadow sets to fall out of steady state. You'll either have toe/ > suffer through a shadow merge or shadow copy.   C Assuming the failed node had the shadowset mounted, a merge will besC triggered.  You could remove all but one shadowset member and startRB shadow copies instead of the merge, but you'd lose your redundancy	 that way.    > If you have a controller that N > supports logging (e.g. HSJ50s with the right firmware), the operating systemN > will often be able to engage in a mini-merge, theoretically lowering the I/O > load of a full merge.u  A Only the areas recently written to by the failed node are merged, C which drastically reduces the amount of area to be merged, and thus-D the duration of the merge operation -- usually down to mere seconds, compared with possibly hours.a  2 > This will only work if the two volumes have been& > created on the same controller pair.  D No, what matters is that all the members are on controllers with theF Volume Shadowing Assist called Write History Logging.  They don't haveE to be on the same controller pair, and in reality one would typically D want to avoid having them on the same controller pair if possible to? prevent a problem in case a controller pair failed to fail-over D properly or a hardware problem in one controller prevented the other from taking over.   A The Volume Shadowing Assist called Disk Copy Data (which can helpnF speed shadow copy operations) is only available if both members are onA the same controller.  Perhaps that's what you were thinking of...a  < > Further, during this time frame, you have no protection inD > case of failure of the volume that is the source of the shadow setI > resolution. This is also the dirty little secret of multi-site clusters , > relying exclusively on software shadowing.  @ This is false.  During a merge operation, data on all members is equally valid.  E Volume Shadowing needs only meet the semantics of an individual disk, F which are that when a VMS system fails, you do not know whether or notC a given write operation really made it to the disk or not, but it'slD important that whichever result occured (write was lost, so old dataE remains on the disk; or write succeeded, so new data is on the disk), B you always get the same result from any subsequent reads after theD failure as you get from the first read.  It doesn't really matter ifC you get the older or the newer data, as long as you don't sometimesAD get the old data and sometimes get the new data when you do a read. B (Folks who really need to guarantee that the in-flight transactionD gets completed (data gets written) or rolled-back (old data remains)F use a journaling product like RMS Journaling, or use a scheme like theC log or journal files that database products use to properly recoverh after a system failure.)  C Shadowing normally keeps all disks identical, and thus can normallyt? read data from any one of the member disks.  But after a systemdD crashes, there might be some blocks different between members, so if@ Shadowing followed its default behavior and satisfied reads fromD different members, it might return different data different times onB read requests satisfied from different members.  So Shadowing mustD take special care during a merge operation to guarantee that it acts4 like (provides the same semantics of) a single disk:A 1) If you read blocks from an area which has not yet been merged,nE Shadowing compares the blocks right then and there, and if there is aiB difference, it arbitrarily picks one (from whichever volume it hasC designated as the shadow master, but which might contain either theeE old or the new data, since there's no guarantee which of the parallelt= writes landed first) and copies that data to the others, thenfC double-checks to make sure it is the same before returning the readh data to the user.lD 2) A merge thread scans across the entire disk, comparing and fixingD up any differences.  Once the merge thread scan completes, Shadowing@ can return to its normal steady-state behavior of directing read+ operations to any of the shadowset members.   L > Controller-based hardware mirroring has great performance characteristics;H > however, I would never count on it unless I was running dual-redundant0 > controllers with battery backup for the cache.  E I'd add that I'd also want mirrored cache, so that failure of a cachenC module would not result in loss of data.  If I didn't have mirroredtB cache, I'd want to shadow across controllers -- I've experienced a< (non-mirrored) cache module failure, and been saved by HBVS.  : > Even normal shutdown of dual-redundant controllers won't > affect the mirrror set.   D But note that a crash of a controller will result in a normalizationC operation (equivalent to a shadow full-copy), since it then has themD same problem to solve as the host-based volume shadowing product has when a host fails.  . > Personnaly, I have never seen a failure of aN > controller-based mirror set without a failure of one of the disk drives that > implements it.  D Wait -- aren't mirrorsets supposed to protect against failure of any one of the disk drives? :-)  M > Another nice thing about controller-based mirror sets is automatic recoverylL > of disk failure. When setting things up, you should place a few extra disk  D I agree.  Sparesets are an area where controllers put HBVS to shame.   ------------------------------    Date: 17 Apr 2003 23:08:12 -0500+ From: young_r@encompasserve.org (Rob Young)B3 Subject: Re: Mirroring versus Shadowing - any diff?A3 Message-ID: <DE$xttkrSTrw@eisner.encompasserve.org>1  b In article <3E9E76FA.429A316C@vl.videotron.ca>, JF Mezei <jfmezei.spamnot@vl.videotron.ca> writes:   > N > Host based shadowing does have a neat capability though: if you have 4 nodesO > in a cluster, with one node doing a hell of a lot of "reads", it can actuallyrH > use MSCP to spread the reads across all nodes giving you built-in loadO > balancing. If you have 4 drives in an array, each node then has a single pathaH > to the array so there is only so many concurrent IOs that can be done. >   F 	You might also want to consider shadowing RAID 0+1 or similar to jack1 	your IO rates (if you can afford to or need to).   N > Another aspect of VMS host based shadowing is that when your write completesM > at the application level, you are garanteed that the write completed on allmK > drives. That si not always the case on a disk array, especially if it haspN > fancy caching/buffering. (I guess that problem can be alleviated with a goodO > UPS system for the disk array to ensure that any data in the buffer does make 
 > it to disk.o  I 	But as the 1990's stretch into the 2000's , many of us are using shadowsgF 	in conjunction with writeback cache at a controller level (1).  Some H 	of us are fortunate enough to have those controllers in different data 	 	centers.a   				Rob    (1)  This isn't new or uncommons   ------------------------------   Date: 18 Apr 03 07:42:34 +0200) From: p_sture@elias.decus.ch (Paul Sture)nN Subject: Re: Personal Firewalls - was Email to Geoff.Graves@hp.com is rejected) Message-ID: <Fn1234baz$ZE@elias.decus.ch>l  j In article <20030414110902.10.qmail@gacracker.org>, Doc.Cypher <Use-Author-Address-Header@[127.1]> writes:@ > On Mon, 14 Apr 2003, Nic Clews <sendspamhere@127.0.0.1> wrote: >>JF Mezei wrote:e >>>  >>> Paul Sture wrote:iM >>> > pretty scathing about "personal firewalls" such as BlackIce Defender ore! >>> > ZoneAlarm. Here's the link:o- >>> > http://www.samspade.org/d/persfire.htmlc8 >>> > and '"Personal Firewalls" are mostly snake-oil' at. >>> > http://www.samspade.org/d/firewalls.html >>> N >>> Is there anyone who really beleived that a piece of windows software could8 >>> ever protect its own instance of a windows machine ? >>= >>A port sniffer type worm or virus is blocked by firewallingd' >>applications, so there is some value.  > K > Since I started using Kerio on my PC, I'd say these software firewalls do D > have some value. Every layer of security you can add helps make anH > intrusion more difficult. Comparing Kerio with ZoneAlarm, I'd call the > latter a toy.e >   C I was hoping someone might jump in with a reference to Kerio, sincec) that's what Huntler Goatley recommends ate  +      http://www.goatley.com/hunter/w2k.htmli  C I somehow suspect that Hunter knows a bit more than me about TCP/IPs- (understatement of the year?), hence my post.l  G >>Sam Spade is scathing from the point of view that these products view J >>all requests as "intrusions" which as it points out is generally not the6 >>case, it only ends up making the user more paranoid. > M > Well, amongst other things, I've been able to lock down Outhouse Express soiD > that it can only access a limited set of ports and addresses. WithF > everything else denied, I get lots of HTML mail which doesn't render > properly.r >   ! That on its own is justification.e   -- I
 Paul Sture   ------------------------------  % Date: Thu, 17 Apr 2003 16:24:51 -0600o$ From: Lee Y T Mah <lytmah@cha.ab.ca>2 Subject: Problem with formfeed on the Genicom 4400) Message-ID: <3E9F29B2.FC926DA2@cha.ab.ca>M  - Require assistance from someone familiar with ) Genicom printing who may have encountereds this same problem.  , On a VAX system I have a LAT queue and an IP- queue going to the same Genicom 4400 printer. * I print the same test file to both queues.. The LAT queue prints thru a port on a terminal4 server and formats a page as 44 lines (as expected).- The IP queue is formatting a page as 88 linesB (incorrectly).1 Both queues are referencing the same form and twon1 library modules for printing.  The 2 modules are:s$   Module GEN4400_44LEN, form length:;     <ESC>[4950 r    (5.5" form length: 55 * 720 / 8 = 4950)1:   Module GEN4400_8_10, eight lines per inch and ten-pitch:     <ESC>[90;72 Ge   The form is MYFORM44:cE Form name                                        Number   Descriptione= -------------                                       --------- 
 -------------e; MYFORM44 (stock=MYFORM)     2345        Special Form (4400) 2     /LENGTH=44 /SETUP=(GEN4400_44LEN,GEN4400_8_10)&     /STOCK=MYFORM /TRUNCATE /WIDTH=160   The IP queue is: $ SHOW QUEUE/ALL/FULL  MYQ99H Printer queue MYQ99, idle, on J::"172.X.Y.Z:9100", mounted form MYFORM44    (stock=MYFORM)    <Genicom 4400><   /BASE_PRIORITY=4 /DEFAULT=(FORM=DEFAULT) /NOENABLE_GENERIC /LIBRARY=GNH$TESTs6   Lowercase /OWNER=[SYSTEM] /PROCESSOR=TCPIP$TELNETSYM0   /PROTECTION=(S:M,O:D,G:R,W) /SCHEDULE=(NOSIZE)  5 The value for the related formfeed system logical is: .     TCPIP$TELNETSYM_SUPPRESS_FORMFEEDS" = "35"   Are my control codes correct?P5 Is the IP symbiont even using the form length module?v  9 Note that the LAT queue references a terminal for output.0) The terminal displays a page as 44 lines.V1 The IP queue has no comparable device for output.t     -- Leeu  5 L Y T Mah                    Email:  lytmah@cha.ab.caM   ------------------------------  % Date: Thu, 17 Apr 2003 20:36:43 -0400N0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>6 Subject: Re: Problem with formfeed on the Genicom 4400/ Message-ID: <3E9F4883.5BC21DBC@vl.videotron.ca>=   Lee Y T Mah wrote:7 > Is the IP symbiont even using the form length module?hL LATSYM and TELNETSYM are two different beasts. There was a recent discussionQ on some of the symbionts not fully supporting all the defaylt VMS print features.h  H You can do a  $PIPE SHOW DEV/FILES/NOSYS DKAxxx: | search sys$input symb  M Look for the .TLB files that are opened. You should see if the telnet and lat<- symbionts use the same device control libraryf  N Note that some symbionts may not keep the device control library opened at all
 times though.d   ------------------------------  % Date: Thu, 17 Apr 2003 21:11:21 -0600i$ From: Lee Y T Mah <lytmah@cha.ab.ca>6 Subject: Re: Problem with formfeed on the Genicom 4400) Message-ID: <3E9F6CD9.F4B594F2@cha.ab.ca>,  > I initially used the same module and form plus the VMS command: DEFINE FORM/LEN=44 to specify the form length for both the? LAT and telnet symbiont queues.  The word from Software Supporti? is that the telnet symbiont does not undertstand/use the LEN=44e@ qualifer.  Therefore I contacted Genicom who sent me the control@ code for 44 lines per page (LPP).  I created a module for 44 LPPF and put it into a test library.  The telnet symbiont is still thinking4 in terms of 88 lines per page, rather than 44 lines.B The report is Fortran control.  I've even created a test file with@ only several lines per record, separated by the Fortran formfeedC "1+".  The IP queue prints several lines, completes a full page forg@ 88 lines, and then continues doing the same with each subsequent record.n     JF Mezei wrote:n   > Lee Y T Mah wrote:9 > > Is the IP symbiont even using the form length module?hN > LATSYM and TELNETSYM are two different beasts. There was a recent discussionS > on some of the symbionts not fully supporting all the defaylt VMS print features.t >oJ > You can do a  $PIPE SHOW DEV/FILES/NOSYS DKAxxx: | search sys$input symb >lO > Look for the .TLB files that are opened. You should see if the telnet and latt/ > symbionts use the same device control library  > P > Note that some symbionts may not keep the device control library opened at all > times though.o   -- Leeb  5 L Y T Mah                    Email:  lytmah@cha.ab.ca  Capital Health Authority Information Systems, RAH CSC Edmonton, Alberta, CANADAs   ------------------------------  % Date: Thu, 17 Apr 2003 23:59:54 -0400 0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>6 Subject: Re: Problem with formfeed on the Genicom 4400/ Message-ID: <3E9F7815.4FB51CE3@vl.videotron.ca>-   Lee Y T Mah wrote:A > LAT and telnet symbiont queues.  The word from Software SupportfA > is that the telnet symbiont does not undertstand/use the LEN=44lB > qualifer.  Therefore I contacted Genicom who sent me the controlB > code for 44 lines per page (LPP).  I created a module for 44 LPPH > and put it into a test library.  The telnet symbiont is still thinking6 > in terms of 88 lines per page, rather than 44 lines.  M Why doN't you simply remove the "lines per page" number (I think set it to 0) M and let the printer deal with it ? It the telnet symbiont doesn't handled the M splitting of pages, insertion of headers etc, then there i no point trying toM! have to know about any page size.d   ------------------------------  % Date: Thu, 17 Apr 2003 22:03:06 -0600r$ From: Lee Y T Mah <lytmah@cha.ab.ca>6 Subject: Re: Problem with formfeed on the Genicom 4400) Message-ID: <3E9F78FA.74393F27@cha.ab.ca>   M This printer is multi-use and must be flexible enough to handle 44, 66, or 88  lines per page.c   JF Mezei wrote:    > Lee Y T Mah wrote:C > > LAT and telnet symbiont queues.  The word from Software SupporteC > > is that the telnet symbiont does not undertstand/use the LEN=44aD > > qualifer.  Therefore I contacted Genicom who sent me the controlD > > code for 44 lines per page (LPP).  I created a module for 44 LPPJ > > and put it into a test library.  The telnet symbiont is still thinking8 > > in terms of 88 lines per page, rather than 44 lines. >iO > Why doN't you simply remove the "lines per page" number (I think set it to 0)xO > and let the printer deal with it ? It the telnet symbiont doesn't handled theiO > splitting of pages, insertion of headers etc, then there i no point trying tov# > have to know about any page size.a   -- Lee   5 L Y T Mah                    Email:  lytmah@cha.ab.cao Capital Health Authority Information Systems, RAH CSC Edmonton, Alberta, CANADA    ------------------------------  # Date: Thu, 17 Apr 2003 23:48:51 GMT-# From: "John Smith" <a@nonymous.com>r Subject: Re: SARS Impact on HPH Message-ID: <D3Hna.71234$Vzu.25641@news02.bloor.is.net.cable.rogers.com>  = "JF Mezei" <jfmezei.spamnot@vl.videotron.ca> wrote in messagee) news:3E9E305F.3F86FC5E@vl.videotron.ca...oC > About 200 Toronto (canada) Area HP employees were quarantined fors about a weekF > because they had come into contact with employees who had shown some SARS symptoms. > C > Ontario Health officials have now made a call for anyone with anyb
 signs of aB > flue with fever to stay at home, in order to prevent such events which could F > cripple some companies. The government has waived a 2 week period to geta9 > unemployment insurance for workers who are quanantined.s > E > I bet the cable/dsl providers will get a lot of orders so that many  workersy > can work from home.r    F http://www.globeandmail.com/servlet/ArticleNews/TPStory/LAC/20030417/R HPHP/TPBusiness/TopStories    How HP handled urgent SARS alert  ! Emergency plan forced into action  By KATHERINE HARDING' With files from reporter Jacquie McNisht    / UPDATED AT 7:43 PM EDT  Thursday, Apr. 17, 2003a  > It took just a single phone call late in the day on April 8 to? Hewlett-Packard (Canada) Ltd.'s manager of health and safety totA trigger a "call down," an emergency alert to contact and lock outt  employees at its Markham office.  F The HP manager had been warned that a probable SARS patient had brokenE a quarantine order by going to work, and the firm suddenly became the.< first in Canada to execute a mass quarantine of employees inB connection with the outbreak of severe acute respiratory syndrome.  E Within hours, 102 employees and 95 recent visitors to the informationcE technology processing facility were told to isolate themselves for 10l? days and wait for a call from the local health unit for further, information.  E "By that time, most people had already left for the day," said AngelaoF Rea, HP Canada's spokeswoman. "We basically went into a 'call down.' "  C A small team of managers contacted every employee who worked at thel> facility. The managers also phoned contract workers and recent@ visitors to the office using a sign-in book that's posted at the reception desk.c  A "You must move quickly, and you must do what the authorities tellr you," Ms. Rea said.   A She added that the Mississauga-based company's detailed emergencyy> contingency plan helped it maintain operations at the affectedC facility throughout the ordeal, while still obeying orders from thea health unit.  0 "Companies have to think about any eventuality."  D Ms. Rea said HP Canada had the plan in place long before SARS poppedC up. "It relates to any kind of 'what if' plan that a company should 
 consider."  C The day after the quarantine order was made, the Markham office was ? staffed by a reduced work force of HP workers culled from otherr  facilities and contract workers.    The facility never had to close.  F To make sure quarantined employees didn't try to go to work, HP Canada= de-activated their security passes. The company also arrangedlA telecommuting for affected employees who volunteered to work fromu home.a  D Ms. Rea said a small team of employees from almost every department,E including communications and human resources, are helping the company < get through the crisis. "It's been a team effort to mobilizeC everything and to make sure we are doing what we are supposed to beo doing."h  D Employee communication is key and the company has been issuing dailyE updates about the situation, Ms. Rea said. "During this occurrence wetA knew we wanted to be open with our employees. We didn't want them F reading the news in the newspaper. We want them to get the information from us so that it's accurate."   F Last Monday, the local health unit told some of the affected HP Canada, employees they could emerge from quarantine.  D The employee at the centre of the mass quarantine order is currentlyB in the hospital with what is considered a probable case of SARS. A> second HP Canada employee from the same facility has also been+ hospitalized with a suspected case of SARS.o  @ York Regional Police launched an investigation into the breachedD quarantine order last week, but Ms. Rea wouldn't comment about aboutA the police involvement. "We can't confirm that indeed that personlD broke quarantine . . . We haven't had an opportunity to talk to them+ because they are sick and in the hospital."   9 HP Canada has recently banned business travel to and fromh= SARS-affected places, including Hong Kong and mainland China.t  A It has also instructed employees returning from personal trips to E those destinations to stay home for 10 days before returning to work.a  D That rule is also being applied to visitors to HP Canada facilities.? Receptionists are asking all visitors whether they've been to aaF SARS-affected country within the past 10 days, and if they answer yes,1 they aren't allowed on to the company's premises.n  E Jim Hassell, a labour and employment law specialist at Osler Hoskin &t? Harcourt, said most employers were complacent about the risk of D business interruption until Hewlett Packard was shut down last week.  D "Hewlett Packard was a big wake-up call. It made people realize thatB one person coming in and breaking quarantine could end up shutting down your entire operations."l   An eight-step contingency plan  C Hewlett-Packard (Canada) moved quickly to deal with the SARS crisisn@ when it had to keep operating during a mass quarantine. Here are lessons for other companies.  F 1. LOCK THE DOORS: Affected employees shouldn't be allowed back on the! premises under any circumstances.n  F 2. PHONE EVERYBODY: Use updated employee contact lists, so that peopleA can be reached after hours. Also use guest sign-in information toi inform recent visitors.k  = 3. ORGANIZE NEW TEAM: Know who is going to take over affectedoF operation. HP Canada used a mixture of employees from other facilities and contract workers.t  A 4. COMMUNICATE: Let employees know what's going on throughout ther4 crisis. Clear up rumours and misconceptions quickly.  F 5. ASSIGN LEADERS: Make sure a team of people are in charge during the< crisis. Try to include representation from every department.  A 6. TELECOMMUTING: HP Canada made alternative work arrangement fore* employees if the wanted to work from home.  ? 7. STAY IN TOUCH: Get the latest information from the local andl3 provincial health authorities as often as possible.a  F 8. HAVE A PLAN: HP Canada saved itself a lot of headaches by having an? emergency plan in place long before the quarantine was ordered.l    > (c) 2003 Bell Globemedia Interactive Inc. All Rights Reserved.   ------------------------------  # Date: Fri, 18 Apr 2003 00:20:08 GMTe+ From: LESLIE@JRLVAX.HOUSTON.RR.COM (leslie)- Subject: Re: SARS Impact on HP< Message-ID: <YwHna.101212$rd4.3970916@twister.austin.rr.com>  1 JF Mezei (jfmezei.spamnot@vl.videotron.ca) wrote:eJ : About 200 Toronto (canada) Area HP employees were quarantined for about H : a week because they had come into contact with employees who had shown : some SARS symptoms.w : H    http://www.siliconindia.com/shownewsdata.asp?newsno=19213&newscat=Top'    First case of SARS reported in Indiaa    e HP has operations in India:n  !    http://www.digitalindiasw.com/     Digital GlobalSofts  /    http://h18001.www1.hp.com/partners/digiglob/     HP and Digital GlobalSoft  : and uses some of those employees in the U.S. on L-1 visas.  2 --Jerry Leslie   (my opinions are strictly my own)9   Note: leslie@jrlvax.houston.rr.com is invalid for email.   ------------------------------    Date: 17 Apr 2003 14:07:51 -0500+ From: young_r@encompasserve.org (Rob Young)(' Subject: Re: StorageTek on SAN with VMSs3 Message-ID: <hILEMsjPTA5F@eisner.encompasserve.org>m  a In article <dedt9vomib7r0q53n3s2lfj885bt6aunc1@4ax.com>, jlsue <jefflsxxxz@sbcglobal.net> writes:e9 > On Tue, 15 Apr 2003 22:05:19 -0500, "David J. Dachtera".  > <djesys.nospam@fsi.net> wrote: >  >>G >>The challenge in this setup is that it's supposed to be an EnterpriseiI >>SAN solution. So, the ACSLS part is there to play server to the various H >>clients wanting robotic actions: W2K, Novell and VMS. We all share theJ >>robots (two ACSs). The drives are dedicated through zoning on the switch- >>and through mapping on the SCSI-FC bridges.S >>' >>This is turning into one unholy mess.n > L > Yes, this is one case where a roll-your-own could be complex, and become a > support/admin's nightmare. > F > And, unfortunately, ABS does not support Novell, and there are "some@ > restrictions" for Windows 2000 clients (accordign to the SPD). >    	[snip]r   > F > Other than all this, I'd look into another EBS product, like HP DataD > Protector, VERITAS NetBackup, or Legato Networker (and potentiallyK > integrate with HP EVM to automate snapshots).  One of these servers wouldsL > manage the entire library and allocate drives to whatever server needs it,3 > as well as keep an on-line catalog of everything.r >   A 	Right.  I was hoping this thread would turn to this.  You really D 	want to budget and buy TSM/ADSM (first choice), Veritas, or Legato.    E 	Attempting to use those drives independent of an EBS is silly/crazy.   @ 	How would it scale?  What about when you are backing up storage= 	associated with 50, 100, 200 servers with SAN and DAS?  (DASt; 	will always be there if for no other reason boot devices).t  C 	You really can scale the EBS in several directions.  For instance:n  V http://www.storserver.com/main.cfm?menu=2&submenu=vbm&detail=include/seriessummary.inc  F 	You could purchase something similar to the high-end solutions there.F 	Fibre attached tape drives, 1.5 TB disk pool.  You could be streamingG 	in several directions (to tape and to disk pool).  There are solutionsoF 	larger than that as you know with the StorageTek tape solutions, they 	aren't small (in most cases).  D 	Point is with a disk pool you can use that to stream high bandwidthE 	to it.  Much higher than you can throw at a tape drive and that willo 	help you meet backup windows.   				Rob-   ------------------------------  % Date: Thu, 17 Apr 2003 16:47:22 -0400F* From: "Syltrem" <syltremzulu@videotron.ca>  Subject: Strange SYS$ENQ problem4 Message-ID: <drEna.4526$945.13337@tor-nn1.netcom.ca>  J I am calling a basic subroutine from a Powerhouse QUICK program (for thoseI who have worked with powerhouse, my subroutine is linked in QKDRIVER.EXE)l   The basic subroutine does a D          STATUS = SYS$ENQW(,LKMODE BY VALUE, CTRL_LKSB(0%) BY REF, &A                            FLAG BY VALUE, TEMPLOCK BY DESC,,,,,,)? among other things.aG The result of the call is 1 (success) and the status in lksb is also 1.e  F What happens is that QUICK exits to DCL immediately after calling this' subroutine (which by the way exits with 3 status = 1 and also does what it's supposed to do).eL If I remove that particular statement from the basic subtoutine, all is well' (except I don't get the result I want!)n   I don't know why it does this.L It is the 1st time this ever happened to me. I call all kind of routines all? the time (from QUICK), and I've never encountered any problems.c  J I traced the QUICK program with the debugger and right after it calls that routine, it goes back to DCL.u  - Can anyone give me a clue of where to search?    --   OpenVMS 7.2-1, Oracle 8.1.6.0e   SyltremtI http://pages.infinit.net/syltrem (OpenVMS related web site - en franais) 8 To reply to myself directly, remove zulu from my address   ------------------------------  % Date: Thu, 17 Apr 2003 20:46:13 -0400l0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>$ Subject: Re: Strange SYS$ENQ problem/ Message-ID: <3E9F4ABD.C10E6721@vl.videotron.ca>e   Syltrem wrote:F >          STATUS = SYS$ENQW(,LKMODE BY VALUE, CTRL_LKSB(0%) BY REF, &C >                            FLAG BY VALUE, TEMPLOCK BY DESC,,,,,,)   I > The result of the call is 1 (success) and the status in lksb is also 1.a > H > What happens is that QUICK exits to DCL immediately after calling this > subroutine  G Is it possible that some other part of the program has some sort of AST C pending that is triggered when the lock you try to get is acquired.r  G Have you tried to request a totally different, totally unrelated lock ?u    N > It is the 1st time this ever happened to me. I call all kind of routines allA > the time (from QUICK), and I've never encountered any problems.   N When you fool around with all kinds of routines all the time, you shouldn't beI surprised if you catch some nasty bug. Haven't your parents ever told youa about safe-programming ?  M Another thing you could try is to $SETAST to OFF prior to thw ENQW, then do aiN few things (including some prompt to continue, and then $SETAST back to ON. IfJ the program quits due to an AST being triggered, that should give you some1 hint. (unless the AST is at a much higher level).:   ------------------------------  # Date: Fri, 18 Apr 2003 04:05:52 GMTd& From: "BC Buyer" <BCbuyer@hotmail.com>$ Subject: Re: Strange SYS$ENQ problem= Message-ID: <AQKna.747169$Yo4.71836138@news1.calgary.shaw.ca>i  5 "Syltrem" <syltremzulu@videotron.ca> wrote in messageh. news:drEna.4526$945.13337@tor-nn1.netcom.ca...L > I am calling a basic subroutine from a Powerhouse QUICK program (for thoseK > who have worked with powerhouse, my subroutine is linked in QKDRIVER.EXE)k >h > The basic subroutine does arF >          STATUS = SYS$ENQW(,LKMODE BY VALUE, CTRL_LKSB(0%) BY REF, &C >                            FLAG BY VALUE, TEMPLOCK BY DESC,,,,,,)n > among other things.nI > The result of the call is 1 (success) and the status in lksb is also 1.    (snip)   >)L > I traced the QUICK program with the debugger and right after it calls that > routine, it goes back to DCL.u >w  K I've had Heisenberg come and bite me in similar circumstances. By running a K problem child program under the debugger, I had programs crash in ways thatlA I didn't expect given the other behavior I observed. In these fewtG circumstances, it turned out that my process was hitting resource quotanL limits that a non-debug run of the program would not reach. In effect, I wasK changing the system under study by observing it! Once this was realized andsL the appropriate quotas raised, the debugger run quickly established the trueI source of the problem in a different area of the program entirely. I haverJ found that this problem will most frequently show itself when a new module
 is scoped in..  K I assume that you traced execution into and through your BASIC routine. Are K you setting FLAG with LCK$M_NOQUEUE? If not, how are you detecting that the  lock has been granted?  B The other way I have seen quick deaths like this is with an accessI violation. You have probably checked already, but you might want to countp> your commas and verify your data types and passing mechanisms.     Cheers,B  Peter Smode   ------------------------------    Date: 17 Apr 2003 12:46:32 -0700% From: tadamsmar@yahoo.com (Tom Adams)h& Subject: system tools cd mount problem= Message-ID: <ea44f5a1.0304171146.6adb364a@posting.google.com>a  A When I try to mount the system tools CD (September 1999) I get a:t  $ MOUNT-F-NONEXPR, nonexistent process  ) error messsage and the CD is not mounted.n  D I tried two different Compaq System Tools CDs from quarterly updates andi6 got the same message.  I mount other CD with problems.    Any ideas why this is happening?  : I used the mount command specified on the front of the CD.  B I am trying to put a version of Compaq Analyze on a system that is running B 7.2-1, so I was going back to an earlier System Tools CD to try to> find a version of Compaq Analyze that is consistent with 7.2-1   ------------------------------  % Date: Fri, 18 Apr 2003 01:45:42 -0400f' From: "Wayne W. Scott" <wscott@nac.net>a/ Subject: tcpip 5.3-1 socket inactivity timeout? & Message-ID: <3E9F9106.9020504@nac.net>  	 Hi Folks:e  P Does TCP/IP 5.3-1 close sockets if there is no activity for some period of time?  Y I am not asking for something that I want to happen; I am trying to debug a condition in  U which it appears that socket connections are being broken unexpectedly and I need to y know what, where, when.d   Thanks,t WayneT   ------------------------------  % Date: Thu, 17 Apr 2003 14:32:12 -0400e0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>/ Subject: Re: TCPIP Services and IP masquerading2/ Message-ID: <3E9EF32B.BF1CF88D@vl.videotron.ca>o   Phillip Helbig wrote:j > > Have you checked the logs ?n >  > What logs?  # SYS$MANAGER:DECW$SERVER_0_ERROR.LOGp  M I think that there is a logical one can set to increase the level of logging.aG But if a call makes it to your machine and is refused by the decwindowsg$ server, it will show up in that log.  J > The same machine CAN locally display remote applications through anotherI > router, so it must be a problem with (the configuration of) the Linksys 	 > router.   L From the outside, what happens if you attempt to telnet to port 6000 ?  ThisM may yield some clues as to where this stops. Does yor router have the abilityeM to send messages via a syslog mechanism ? I have that on my netgear and foundeK this to be extremely useful. (I hear the call coming on because of the beepl  generated by the opcom message).   ------------------------------    Date: 17 Apr 2003 15:48:05 -0500; From: koehler@eisner.nospam.encompasserve.org (Bob Koehler)mD Subject: Re: Truly effective - let's get them for the VMS campaign!!3 Message-ID: <7yzBfJrh8QRL@eisner.encompasserve.org>t  n In article <erxna.66170$Vzu.11664@news02.bloor.is.net.cable.rogers.com>, "John Smith" <a@nonymous.com> writes:   >             Low Score Wins >  > OpenVMS                    x > Unix (all vendors)          xi$ > Linux                            x > Microsoft Windows      xA > (all numbers cumulative lifetime incidents for all commerciallye > available versions)t  0    I hope that lost something in the formatting.   ------------------------------  # Date: Thu, 17 Apr 2003 23:54:50 GMTl# From: "John Smith" <a@nonymous.com>tD Subject: Re: Truly effective - let's get them for the VMS campaign!!G Message-ID: <e9Hna.71252$Vzu.7217@news02.bloor.is.net.cable.rogers.com>i  @ "Bob Kaplow" <kaplow_r@eisner.encompasserve.org.TRABoD> wrote in5 message news:7xAVMvnGdSPU@eisner.encompasserve.org...h > In articleC <erxna.66170$Vzu.11664@news02.bloor.is.net.cable.rogers.com>, "Johnn Smith" <a@nonymous.com> writes:o? > > Imagine a company like Honda, clearly an engineering driven 8 > > organization with a flair for marketing, owning VMS. > >r > > drifting.... > >rE > > How about a print ad that reads like the following where x is thei > > actual value for each: > >  > >             Low Score Wins > >S  > > OpenVMS                    x! > > Unix (all vendors)          x & > > Linux                            x > > Microsoft Windows      xC > > (all numbers cumulative lifetime incidents for all commerciallyo > > available versions)l > >uC > > Your company may never have heard of CERT* but your bottom linee hass" > > certainly been affected by it. >s@ > Hey, Sue! You should get the CERT numbers for all of these and
 actually fill A > this in and make it public. Break down the UNIX by vendor. Yea,t you're gonna3 > have to list Tru64 and HPUX. And include NSK too.d >m; > And send "John Smith" a check. Or at least a VMS t-shirt.l    A I'd rather HP kept the money the t-shirt cost and put it into theaF OpenVMS Advertising Fund jar that HP keeps in the executive snack areaF in Palo Alto for people to put their spare change into. I hear it's up to about $17.50 so far.    ------------------------------    Date: 17 Apr 2003 12:47:34 -0700+ From: hobbesnet@hotmail.com (Scott Squires)l) Subject: Re: VAX C++ 5.6 install questions= Message-ID: <ff921edf.0304171147.5a1b182b@posting.google.com>i  ] John Santos <JOHN@egh.com> wrote in message news:<1030416235439.2426B-100000@Ives.egh.com>... & > On 15 Apr 2003, Scott Squires wrote: >  > > ...fE > > $ @sys$update:vmsinstal cxx056 disk$user:[programs.cxx] options ld > = > This is an ancient version of C++.  Current version is 6.5.  > ... 0 > Are you sure C++ 5.6 is supported on VMS V7.2? > ...:C > Hmmm???  It's mucking around with systartup_V5.  This implies VMS D > V5.x, since this file's name changed to systartup_vms.com at V6.0.  D I don't know why it is doing that; the comments in the kitinstal.com mention VMS 6.2.  / > Try installing a more-current version of C++.l  C I was not aware there was a newer version for VAX.  I obtained thisrA kit from ftp://ftp.compaq.com/pub/products/C-CXX/openvms/cxx/.  Ie@ would be enthused if there was a newer version, because at first> glance it appears that this version's STL might not follow theE standard.  However, as my VMS use is a nostalgic pass-time, I have no D budget to buy media -- I have to manage with what is on the hobbyist  cd and what I can find for free.   Regards,
 Scott Squires    ------------------------------  % Date: Thu, 17 Apr 2003 16:22:53 -0400t0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>) Subject: Re: VAX C++ 5.6 install question / Message-ID: <3E9F0D15.B0594725@vl.videotron.ca>t  K I knew I had some memory of problems installing the C++. I recall having tosM check individual includes because C++ installed .h files above those of DEC Co0 and I have to decide which were more up to date.  M But now that you mentioned debugshr, I too had that problem. Unfortunatly, itl9 hadn't popped up initially when i did a serach of google.   H use google to search for : C++ debugshr group:comp.os.vms  You will find1 messages from Steven M. Schweda  (sms@something) e  M Another thread (that spawn the above one) can be found by searchiong for C++   DCPS group:comp.os.vms  M There are a few more things to watch for towards the end of the installation.u   ------------------------------    Date: 17 Apr 2003 18:51:49 -0700+ From: hobbesnet@hotmail.com (Scott Squires)s) Subject: Re: VAX C++ 5.6 install question:< Message-ID: <ff921edf.0304171751.56cf72a@posting.google.com>  g JF Mezei <jfmezei.spamnot@vl.videotron.ca> wrote in message news:<3E9F0D15.B0594725@vl.videotron.ca>...gM > I knew I had some memory of problems installing the C++. I recall having to-O > check individual includes because C++ installed .h files above those of DEC Cv2 > and I have to decide which were more up to date.  D I read that thread a couple days ago, where you decided to stay withB the DEC C headers because the C++ installer was unreliable.  I didC make sure to tell it not to purge the old files, so I'll see how iteD goes with the C++ headers and if it doesn't work out I can trash the	 new ones.d  O > But now that you mentioned debugshr, I too had that problem. Unfortunatly, its; > hadn't popped up initially when i did a serach of google.o > J > use google to search for : C++ debugshr group:comp.os.vms  You will find3 > messages from Steven M. Schweda  (sms@something) n  B David Sneddon pointed out that DEBUGSHR is defined /exec, and so IB modified the kitinstal to deassign/exec.  That fixed that problem.  O > There are a few more things to watch for towards the end of the installation.t  B The other problem I had was that it attempted to delete help filesF which didn't exist ([syshlp.cxx$help]*.*, presumably this procedure is, supposed to install them, not delete them!).  D I fixed that, and then it went through ok.  Except that I don't haveA the CXXLSTL help which the CXX help says should be available.  Ito" isn't in the kit either.  Oh well.  E I wrote up a quick test program and it compiled and ran perfectly, sok I'm happy for now.  :-)v  E Now I just have to update my custom dcltables.exe to include C++... IdE guess I'll go back to the regular dcltables until I get around to it.o  B Thanks for everyone's help, I thought I would never get that thing working!
 Scott Squiresu   ------------------------------  % Date: Thu, 17 Apr 2003 22:42:53 -0400a0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>) Subject: Re: VAX C++ 5.6 install question9/ Message-ID: <3E9F660D.2CB8560F@vl.videotron.ca>r   Scott Squires wrote:D > Thanks for everyone's help, I thought I would never get that thing
 > working!  J The big question is why Digital allow that kit out with such poor testing.   ------------------------------  # Date: Thu, 17 Apr 2003 23:38:19 GMT 6 From: peter@langstoeger.at (Peter 'EPLAN' LANGSTOEGER)& Subject: Re: VAX/Alpha VMS differences3 Message-ID: <LVGna.100771$UR.868224@news.chello.at>e  w In article <OFC8E9212E.5CB2D9C8-ON07256D02.0052972D@rsc.raytheon.com>, "David D Miller" <ddmiller@raytheon.com> writes:i# >   SHOW MEMORY due to architecture> >   ODS-5 only on AlphaoA >   SDA has different interface and commands due to VMS structuree6 >   @AUTOGEN/SYSGEN/SYSMAN PARAM due to VMS structure. >a: >Others?  Please send me a list and I'll make a composite.  ! Here is my (very incomplete) list    CISC vs RISC 32 vs 64 bit+ pages vs. pagelets (512 vs 8192 Byte pages)e3 Macro32 vs Macro64 Assembler (and Macro32 Compiler)j SYSGEN vs SYSMAN IOc VIOC vs XFC (and VIOC) C Device Drivers% Partitioning (H/W and S/W aka Galaxy). Hardlinks & Accessdatesm IEEE# RECALL List (20 Entries vs umpteen) 
 RECALL/SEARCHe ANALYZE/IMAGE/SELECTF Modern Hardware (Fast Eth, GbE, ATM, FibreChannel, MemoryChannel, ...) RAD Support 9 VMS Installation/Updates/ECOs (VMSINSTAL vs. CD-ROM/PCSI)n# Standalone Backup vs Standalone VMSy CLUE.EXE vs SDA>CLUE LINK/SYSEXEa JAVA MOZILLAo APACHE (PERL, PHP, TOMCAT) SSL  CDSAC DECwindows-MOTIF Common Desktop Environment (CDE aka "New Desktop")a MMOV BIND9  PPPi SSHe F90p DNVRTG vs DNVEXT No GAP No DDCMP AdvancedServer AMDS vs AVAILMAN (JAVA)h  OpenVMS Management Station (TNT) GTK+ CXML   Please post your final list ;-)I   -- S Peter "EPLAN" LANGSTOEGER-% Network and OpenVMS system specialist: E-mail  peter@langstoeger.atF A-1030 VIENNA  AUSTRIA              I'm not a pessimist, I'm a realist   ------------------------------    Date: 17 Apr 2003 19:34:48 -0500- From: Kilgallen@SpamCop.net (Larry Kilgallen)m& Subject: Re: VAX/Alpha VMS differences3 Message-ID: <FV595DV5ueve@eisner.encompasserve.org>u  l In article <LVGna.100771$UR.868224@news.chello.at>, peter@langstoeger.at (Peter 'EPLAN' LANGSTOEGER) writes:y > In article <OFC8E9212E.5CB2D9C8-ON07256D02.0052972D@rsc.raytheon.com>, "David D Miller" <ddmiller@raytheon.com> writes:B$ >>   SHOW MEMORY due to architecture >>   ODS-5 only on AlphaB >>   SDA has different interface and commands due to VMS structure7 >>   @AUTOGEN/SYSGEN/SYSMAN PARAM due to VMS structure.m >>; >>Others?  Please send me a list and I'll make a composite.s > # > Here is my (very incomplete) list  >  > CISC vs RISC > 32 vs 64 bit- > pages vs. pagelets (512 vs 8192 Byte pages)e5 > Macro32 vs Macro64 Assembler (and Macro32 Compiler)  > SYSGEN vs SYSMAN IO  > VIOC vs XFC (and VIOC) > C Device Drivers' > Partitioning (H/W and S/W aka Galaxy)N > Hardlinks & AccessdatesN > IEEE% > RECALL List (20 Entries vs umpteen)D > RECALL/SEARCHq > ANALYZE/IMAGE/SELECTH > Modern Hardware (Fast Eth, GbE, ATM, FibreChannel, MemoryChannel, ...)
 > RAD Support ; > VMS Installation/Updates/ECOs (VMSINSTAL vs. CD-ROM/PCSI)A% > Standalone Backup vs Standalone VMSL > CLUE.EXE vs SDA>CLUE
 > LINK/SYSEXEY > JAVA	 > MOZILLAP > APACHE (PERL, PHP, TOMCAT) > SSLG > CDSAE > DECwindows-MOTIF Common Desktop Environment (CDE aka "New Desktop")s > MMOV > BIND9N > PPPU > SSHF > F90S > DNVRTG vs DNVEXT > No GAP
 > No DDCMP > AdvancedServer > AMDS vs AVAILMAN (JAVA)t" > OpenVMS Management Station (TNT) > GTK+ > CXML  A Alpha has no SCAN compiler, and if someone built a SCAN compiler, $ Alpha has no Debug support for SCAN.   ------------------------------  % Date: Thu, 17 Apr 2003 21:02:46 -0400b0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>& Subject: Re: VAX/Alpha VMS differences/ Message-ID: <3E9F4E9D.AA798F76@vl.videotron.ca>:    Peter 'EPLAN' LANGSTOEGER wrote:% > RECALL List (20 Entries vs umpteen)D  & VAX VMS 7.* has the big recall buffer.   > RECALL/SEARCHo  J VAX will have it once we can find Guy Peleg, corner him in a dark alley to4 force him to make the patch available on vax 7.2 :-)  ; > VMS Installation/Updates/ECOs (VMSINSTAL vs. CD-ROM/PCSI)r    VAX has both VMSINSTAL and PCSI.  % > Standalone Backup vs Standalone VMSy  L Isn't it a case of Alpha not having standalone backup and both vac and alpha, having ability to boot minimal VMS from CD ?   > APACHE (PERL, PHP, TOMCAT)   Perls runs on vax.   > SSLr   openSSL runs on vax.   > BIND9u  ? Isn't the BIND the same for TCPIP 5.3 for both vax and alpha ? :   > PPP0   Real bummer that vax lacks PPP.b   > No GAP   Is "Gap" a VMS customer ?o  
 > No DDCMP  " No message router on alpha either.   > CXML  J Since source come for the XML is available, that isn't a hindrance to VAX.   ------------------------------  % Date: Thu, 17 Apr 2003 14:42:56 -0400u0 From: JF Mezei <jfmezei.spamnot@vl.videotron.ca>> Subject: Re: VMS to be present at european security conference/ Message-ID: <3E9EF5AF.29F9CB12@vl.videotron.ca>c   Alan Adams wrote:iO > > 2 weeks to go before InfosecurityEurope 2003 opens its doors once again toy  < > Now if they'd thought to mention which country it is in...    L They did: It is in Europe :-) :-) :-) :-)  From their text, one would assume9 that such conference would be well known all over europe.b  J But no matter where it is , it is good to see some of the VMS folks (well,I they were ALL-IN-1 until ALL-IN-1 was moved to Bangalore for maintenance)q participate in such events.w   ------------------------------  % Date: Thu, 17 Apr 2003 23:13:54 +0100l9 From: Alan Adams <alan.adams@orchard-way.freeserve.co.uk>a> Subject: Re: VMS to be present at european security conference? Message-ID: <a8b3c4e44b.Alan.Adams@orchard-way.freeserve.co.uk>o  . In message <3E9EF5AF.29F9CB12@vl.videotron.ca>;           JF Mezei <jfmezei.spamnot@vl.videotron.ca> wrote:o   > Alan Adams wrote:fQ > > > 2 weeks to go before InfosecurityEurope 2003 opens its doors once again tot > > > > Now if they'd thought to mention which country it is in... >  > N > They did: It is in Europe :-) :-) :-) :-)  From their text, one would assume; > that such conference would be well known all over europe.  >   J That identifies the approximate area, but not even which landmass. BritainD is part of Europe, but it's b****dy expensive to drive from there toJ anywhere else in Europe. (At least since Sept 11th it's been a lot cheaper to fly.)  K The first announcement I saw was from a .co.uk address, suggesting Britain,T but I'm still not sure.p  L > But no matter where it is , it is good to see some of the VMS folks (well,K > they were ALL-IN-1 until ALL-IN-1 was moved to Bangalore for maintenance)i > participate in such events.a   --  
 Alan Adams& alan.adams@orchard-way.freeserve.co.uk http://www.nckc.org.uk/h   ------------------------------    Date: 17 Apr 2003 18:41:55 -07001 From: keithparris_NOSPAM@yahoo.com (Keith Parris) > Subject: Re: VMS to be present at european security conference= Message-ID: <cf15391e.0304171741.45f88eeb@posting.google.com>e   Alan Adams <alan.adams@orchard-way.freeserve.co.uk> wrote in message news:<2735a3e44b.Alan.Adams@orchard-way.freeserve.co.uk>...< > Now if they'd thought to mention which country it is in...  L A quick google.com search indicates it's in London, April 29 to May 1.  See 0 http://www.infosec.co.uk/page.cfm/NewSection=Yes   ------------------------------  % Date: Thu, 17 Apr 2003 15:17:06 -0400a* From: "Bill Todd" <billtodd@metrocast.net>" Subject: Re: What is a VMS Cluster2 Message-ID: <vKWcnSSjO5IuYAOjXTWcqw@metrocast.net>  2 "Milton" <mbhewitt@optonline.net> wrote in message2 news:79ft9v8g52sfu06f487sed3pbfhg5rjpe1@4ax.com...   ...e  # > GPFS is also available for Linux: B > http://www-1.ibm.com/servers/eserver/clusters/software/gpfs.html >l& > And has been since November 13, 2001 >iL http://www-916.ibm.com/press/prnews.nsf/jan/4F59645B54FCF0D485256B0300512B03  1 I knew JFS got ported early, but had missed GPFS.k   >iC > There's also a new linux storage and file system architecture and A > implementation, suitable for very large clusters, called LustrecE > A next-generation cluster file system which can serve clusters withaJ > 10,000's of nodes, petabytes of storage, move 100's of GB/sec with state4 > of the art security and management infrastructure. > http://lustre.org/  I A potentially very important product, but a long way from completion yet.    > D > >and I'm pretty sure that Sun now does as well (though it's been aI > >while since I checked and back then there wasn't sufficiently-detailede) > >information to be absolutely certain).e > F > Sun Cluster 3.0 has a cluster file system is based on the proxy file > system pxfs.= > http://docs.sun.com/db/doc/806-1424/6jan9i70j?q=pxfs&a=viewy >d@ > There are also Sun shared filesystems QFS, SAM-FS, and SAM-QFS > http://docs.sun.com/?q=QFS  # Thanks - I'll try to get caught up.t   - bill   ------------------------------    Date: 17 Apr 2003 19:06:15 -07001 From: keithparris_NOSPAM@yahoo.com (Keith Parris) " Subject: Re: What is a VMS Cluster= Message-ID: <cf15391e.0304171806.73da807a@posting.google.com>t  e bill@cs.uofs.edu (Bill Gunshannon) wrote in message news:<b7jpj0$1midr$1@ID-135708.news.dfncis.de>...fI > And DLM is available for AIX and Linux (that I know of, maybe more) and/H > with the release of the code by IBM, on every other Unix very shortly.  D I tracked down that reference when I first heard of it quite a while3 back, but was a bit disappointed with what I found.M  E I seem to recall hearing the AIX version was done so Oracle OPS couldn run in an HA-CMP configuration.a  C For Linux, I presume you're talking about the Linux port of the AIX D HA-CMP DLM code (http://oss.software.ibm.com/dlm/).  It appears thisE code was released in December 2001 in a package called "DLM-Unstable"aA and version 1.0-rc2 (Release Candidate 2, perhaps?).  So the code @ available appears to be of beta quality at best.  Note "The codeD builds and has been tested on 2 node clusters, supporting relatively; light locking loads" but "isn't yet at a production level".i  D The project status was last updated in October 2001.  A quick GoogleF search would make one suspect the project has been dormant since then,> or at least it isn't apparent there's been any rush to port itE anywhere else.  And wouldn't GPL rules make it complex to incorporatem: the Linux code into any proprietary Unix code base anyway?   ------------------------------   End of INFO-VAX 2003.213 ************************