1 INFO-VAX	Sun, 07 Nov 2004	Volume 2004 : Issue 618       Contents:( Re: Ball-park figure for OpenVMS costing( Re: Ball-park figure for OpenVMS costing- Re: how to implement an append only log file? ! Re: LNM$SYSCLUSTER_TABLE question ! Re: LNM$SYSCLUSTER_TABLE question ! Re: LNM$SYSCLUSTER_TABLE question  Re: Performance Tuning Re: Performance Tuning' Re: Request for feedback - COPY utility ' Re: Request for feedback - COPY utility P Re: Sys$Startup:License_Check.Exe (revisited) - OPENVMS-ALPHA not authorized on   F ----------------------------------------------------------------------  $ Date: Sat, 6 Nov 2004 13:50:26 -0500# From: "John Smith" <a@nonymous.com> 1 Subject: Re: Ball-park figure for OpenVMS costing , Message-ID: <tYKdnWqO5IZngRDcRVn-jA@igs.net>  / Phillip Helbig---remove CLOTHES to reply wrote: 5 > In article <clh28b$f1n$1@news.netkonect.net>, Garry + > <asdfasdfadsf@asdfasdfasdfre.com> writes:  > G >> I'm considering using OpenVMS for a web server, it'll be for a small A >> commercial company, so I'm not sure if I can get away with the  >> Hobbyist licence. >  > You can't and you shouldn't. > ? >> Right now I don't want to sign my soul over to HP just for a G >> pricing quote, so can someone give me a ball-park figure for OpenVMS F >> on a single processor 1U rackmount? Something like a DS10L, or if IE >> can get one cheap, a supported Itanium machine. I figure I can get B >> the machine off eBay, so it's just the OS I need the quote for. > H > Consider that something like a VAXstation 4000 might be more than fastE > enough for your needs.  The excellent OSU server by Dave Jones runs @ > fine on VAX.  Or do you need new hardware so you have support?C > Consider getting a dozen VAXstations, which you should be able to D > find for next to nothing; if one goes, just replace the whole box.D > (I'm assuming you would be running the web server in a cluster, so5 > the customers wouldn't notice if a node goes down.)   H For some it may only be worth it if the commercial VMS licences transfer< with the machines, even if the machines themselves are free.   ------------------------------  % Date: Sat, 06 Nov 2004 22:37:36 +0000 - From: Garry <asdfasdfadsf@asdfasdfasdfre.com> 1 Subject: Re: Ball-park figure for OpenVMS costing - Message-ID: <cmjjng$lss$1@news.netkonect.net>   / Phillip Helbig---remove CLOTHES to reply wrote: 5 > In article <clh28b$f1n$1@news.netkonect.net>, Garry , > <asdfasdfadsf@asdfasdfasdfre.com> writes:  >  > H >>	I'm considering using OpenVMS for a web server, it'll be for a small J >>commercial company, so I'm not sure if I can get away with the Hobbyist  >>licence.   >  >  > You can't and you shouldn't. >  >   I I've been conversing with a chap from HP, he has been quite helpful with  # regard to getting a cheap solution.   ? >>Right now I don't want to sign my soul over to HP just for a  J >>pricing quote, so can someone give me a ball-park figure for OpenVMS on K >>a single processor 1U rackmount? Something like a DS10L, or if I can get  I >>one cheap, a supported Itanium machine. I figure I can get the machine  5 >>off eBay, so it's just the OS I need the quote for.  >  > I > Consider that something like a VAXstation 4000 might be more than fast  K > enough for your needs.  The excellent OSU server by Dave Jones runs fine  F > on VAX.  Or do you need new hardware so you have support?  Consider I > getting a dozen VAXstations, which you should be able to find for next  J > to nothing; if one goes, just replace the whole box.  (I'm assuming you J > would be running the web server in a cluster, so the customers wouldn't  > notice if a node goes down.) >   H The machine will go in a rented rackspace company, so has to be 1U, and D has to be reliable. VAX does not interest me, in fact I was kind of D hoping to go right onto IA64 and skip Alpha. I would not be using a 8 cluster, I'm only paying for a single unit of rackspace.  F For the moment, OpenVMS is not going to be cost-effective for me, but G I'm not ruling it out for future projects, which will have security as  7 no.1 priority (cost is no.1 priority for this project).    Cheers   Garry    ------------------------------  % Date: Sat, 06 Nov 2004 23:33:53 -0500 " From: Glenn Everhart <gce@gce.com>6 Subject: Re: how to implement an append only log file?& Message-ID: <418DA5B1.3040002@gce.com>  H Many years ago I wrote a host program for my host-process-as-disk driverK (which has been named various things like frdriver, zrdriver...) that would I act as a write once device above a certain LBN (doing this by reading the I block, checking for a fiducial pattern on it, and allowing overwrite only D if the underlying store had the fiducial pattern). It encrypted dataI written (to defend the abstraction) and used some command files I devised K and init/index=beg/head=nnn and some command files to create the index file I and directories below the fence LBN. (There were later some variants with F FDT intercepts to ensure directories never filled, since they would beE reallocated). Directories had to be preallocated bigger than default.   H This beast was / is kind of a pain to set up because you have to predictE how big directories will be, but it can be used to hold log files and I encrypt them and make sure they cannot be overwritten. The FDT intercepts I also block deletion on the device and certain kinds of writing over files I that are already there, and the block write check absolutely protects any  writing in place.   G This did work in tests with random log files in the sense that they got G created right, wrote to disk only once, and worked completely normally, J except of course they could not be altered. Admittedly the pointers to theH files could be trashed, but the data would still be there immutably, andI because the underlying disk encrypted the results, it was not possible to J simple clobber the container file and not make it obvious someone had beenJ corrupting log files. You could corrupt a file on such a thing, but it wasK pretty hard to edit it so that a change could be hidden. I was well pleased H that the write-once driver level check did not prevent normal sequentialJ writing or appending, and that this function could be so easily had, i.e.,L by pointing the logging to the software "worm" disk. I believe this has beenN on the SIG tapes, since the work was done in the early 1990s as memory serves.  M I don't recall whether I ever altered the host file for alpha, but the driver H works for vax or alpha and has been available in source forever, and theH host process for vax or alpha with that driver is practically identical;J main difference being that the driver UCB definition needs to be right forM the architecture and match the driver used. One simple block cut and paste...   @ It's not such a bad way to do encrypted logs once you get it set@ up...just not so easy to use for general storage as noted above.  J In fact it is about as close to a foolproof system of its type as anything I've ever heard of.    Glenn C. Everhart      Mike Rechtman wrote: > Josef Jarousse wrote:  >  >> Hi, >>  >> I have the following problem: >>H >> I design a system and I would like it to trace each action the systemF >> does. Especially it would be important that even the programm callsH >> from the admin are logged too. What I want is a log file which has anF >> owner that is different from the admin (internal audit for example)F >> and in which the application writes some informations about startedI >> modules etc.. This log file should be "append-only" so that nobody can  >> delete some entries in it. / >> Is there a way to implement such a log file?  >>	 >> Thanks  >> Jo  >  > Start with > $ HELP AUDIT >  > two gotchas:J > 1. There cannot be, by definition, anything not accessible via VMS to a D > sufficiently privileged user. Any logging you start can always be I > stopped, permanently or temporarily by the admin. (Under VMS generally  7 > called the system manager) - thats one of his duties.  > K > 2. Logging everything has the unfortunate result of very quickly filling  F > up any disk, leading either to loss of data (earliest, latest - you  > choose) or some type of DOS. > G > When you write "program calls": is that individual "run" commands or  I > their equivalent (foreign commands, DCL commands) or do you need to go   > to the system-call level?  >  > Mike >    ------------------------------  % Date: Sat, 06 Nov 2004 16:01:51 -0600 2 From: David J Dachtera <djesys.nospam@comcast.net>* Subject: Re: LNM$SYSCLUSTER_TABLE question+ Message-ID: <418D49CE.6FFBFCC8@comcast.net>   / Phillip Helbig---remove CLOTHES to reply wrote:  > / > In article <sxGs9xan6lld@cuebid.zko.dec.com>, 3 > brooks@cuebid.zko.dec.nospam (Rob Brooks) writes:  > G > > When a clusterwide logical is defined, the definition is replicated M > > across the cluster immediately.  That means that each node in the cluster I > > has a local definition of that logical name, such that the definition H > > will survive a rolling reboot.  The only way to remove a clusterwideG > > logical name definition is to reboot the entire cluster at once (or " > > use the DCL command DEASSIGN). > " > Almost.  Consider the following: >  > $ sh log testtest : > %SHOW-S-NOTRAN, no translation for logical name TESTTEST0 > $ define/log/table=lnm$syscluster testtest foo0 > $ define/log/table=lnm$syscluster testtest foo0 > $ define/log/table=lnm$syscluster testtest barB > %DCL-I-SUPERSEDE, previous value of TESTTEST has been superseded- > $ define/log/table=lnm$process testtest foo - > $ define/log/table=lnm$process testtest foo B > %DCL-I-SUPERSEDE, previous value of TESTTEST has been superseded- > $ define/log/table=lnm$process testtest bar B > %DCL-I-SUPERSEDE, previous value of TESTTEST has been superseded > H > This looks to me like, in the case of lnm$syscluster, first a check isG > done to see if the value has changed, but not for other tables.  This H > makes sense, since probably for other tables it is quicker to redefineF > it than to test it, but the opposite is true for lnm$syscluster.  Of# > course, the results are the same.   G I think what must be remembered is that the propagation of cluster-wide G logical names from the DEFINE-ing node to the other cluster partners is D going to take a little time. I'm sure it's possible that you hit theC thing fast enough that it hadn't fully propagated yet. You probably D wouldn't want the DEFINE to be synchronous, and not return until allF nodes have propagated - that could take a while, and if one node has aH problem, it would hang your procedure just waiting for that to complete.  H Even the startup template proc.'s (SY*.TEMPLATE) contain comments notingD that startup-time operations based on cluster-wide logical names mayF need to WAIT until the initialization and propagation of the table has
 completed.   --   David J Dachtera dba DJE Systems  http://www.djesys.com/  ) Unofficial OpenVMS Hobbyist Support Page: " http://www.djesys.com/vms/support/  ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/    ------------------------------   Date: 7 Nov 2004 00:26:41 -0500 / From: brooks@cuebid.zko.dec.nospam (Rob Brooks) * Subject: Re: LNM$SYSCLUSTER_TABLE question- Message-ID: <WMiBbeS0ua74@cuebid.zko.dec.com>   4 David J Dachtera <djesys.nospam@comcast.net> writes:1 > Phillip Helbig---remove CLOTHES to reply wrote:  >>  4 >> brooks@cuebid.zko.dec.nospam (Rob Brooks) writes: >>  H >> > When a clusterwide logical is defined, the definition is replicatedN >> > across the cluster immediately.  That means that each node in the clusterJ >> > has a local definition of that logical name, such that the definitionI >> > will survive a rolling reboot.  The only way to remove a clusterwide H >> > logical name definition is to reboot the entire cluster at once (or# >> > use the DCL command DEASSIGN).  >>  # >> Almost.  Consider the following:   5 	OK, you've got me.  What in the above is "almost"?     I > I think what must be remembered is that the propagation of cluster-wide I > logical names from the DEFINE-ing node to the other cluster partners is F > going to take a little time. I'm sure it's possible that you hit theE > thing fast enough that it hadn't fully propagated yet. You probably F > wouldn't want the DEFINE to be synchronous, and not return until allH > nodes have propagated - that could take a while, and if one node has aJ > problem, it would hang your procedure just waiting for that to complete.  K Well, actually, it *is* synchronous.  (I'm not going to swear to that right L now -- I'm on vacation in lovely Montreal, and don't feel like tracking thisC down right now).  However, there is plenty of defensive code in the K clusterwide logical propagation that will force a crash if an inconsistency 	 is noted.   J > Even the startup template proc.'s (SY*.TEMPLATE) contain comments notingF > that startup-time operations based on cluster-wide logical names mayH > need to WAIT until the initialization and propagation of the table has > completed.  G I think you are confusing two different issues.  At system startup, the O clusterwide logical database is replicated pretty early in the booting process. F If all goes well, it has completed way before any user-written commandH procedures will execute.  However, there is a $GETSYI item code (I thinkM it's CWLOGICALS) that will return TRUE once the clusterwide logical subsystem K has completed its setup.  This replication is far different from that which G takes place when a single clusterwide logical is defined.  In the first I case, there could be many, many clusterwide logicals to be replicated at  G system startup (especially if Host-Based Minimerge is used -- HBMM uses H clusterwide logicals to replicate HBMM policies on a cluster -- hence myI familiarity with clusterwide logicals).  So, before any command procedure G attempts to define or translate a clusterwide logical logical at system K startup, we suggest you verify that F$GETSYI ( "CWLOGICALS" ) returns TRUE. H In fact, some initialization code for HBMM (as part of the SHADOW_SERVERN process) does wait until clusterwide logicals are available before continuing.K (For those keeping score at home, we ran into a race condition for shadowed H system disks for satellite nodes; if an HBMM policy was defined for thatI device, the clusterwide logical name subsystem wasn't always ready before G the SHADOW_SERVER process was up and running, so we needed to check for D CWLOGICALS to be true before resuming SHADOW_SERVER initializion).  D However, once CWLOGICALS returns TRUE, any clusterwide definition isH synchronous with respect to the DCL DEFINE (or ASSIGN) command returning to the prompt.  L This long-winded explanation is likely due to the fact that I've been out ofK the office for several days, have missed talking and thinking about VMS (my M wife, while a fine companion, does not understand any VMS internals) and have P enjoyed many fine meals (with fine wine) in vieux Montreal.  But I digress . . .  K In the second case (the simple defining of a single clusterwide logical via J $CRELNM (which DCL DEFINE ultimately uses, of course)], the replication isN synchronous; the $CRELNM call does not complete until all of the SCS messagingF needed to deal with replication has completed.  If a node is unable toM replicate the logical name, it will die on its sword and crash.  The point is O to guarantee that the logical name is faithfully replicated across the cluster.    --    M Rob Brooks    VMS Engineering -- I/O Exec Group     brooks!cuebid.zko.dec.com    ------------------------------  % Date: Sun, 07 Nov 2004 00:31:31 -0600 2 From: David J Dachtera <djesys.nospam@comcast.net>* Subject: Re: LNM$SYSCLUSTER_TABLE question+ Message-ID: <418DC143.B8AD77A5@comcast.net>    Rob Brooks wrote:  > 6 > David J Dachtera <djesys.nospam@comcast.net> writes:3 > > Phillip Helbig---remove CLOTHES to reply wrote:  > >>6 > >> brooks@cuebid.zko.dec.nospam (Rob Brooks) writes: > >>J > >> > When a clusterwide logical is defined, the definition is replicatedP > >> > across the cluster immediately.  That means that each node in the clusterL > >> > has a local definition of that logical name, such that the definitionK > >> > will survive a rolling reboot.  The only way to remove a clusterwide J > >> > logical name definition is to reboot the entire cluster at once (or% > >> > use the DCL command DEASSIGN).  > >>% > >> Almost.  Consider the following:  > < >         OK, you've got me.  What in the above is "almost"? > K > > I think what must be remembered is that the propagation of cluster-wide K > > logical names from the DEFINE-ing node to the other cluster partners is H > > going to take a little time. I'm sure it's possible that you hit theG > > thing fast enough that it hadn't fully propagated yet. You probably H > > wouldn't want the DEFINE to be synchronous, and not return until allJ > > nodes have propagated - that could take a while, and if one node has aL > > problem, it would hang your procedure just waiting for that to complete. > M > Well, actually, it *is* synchronous.  (I'm not going to swear to that right N > now -- I'm on vacation in lovely Montreal, and don't feel like tracking thisE > down right now).  However, there is plenty of defensive code in the M > clusterwide logical propagation that will force a crash if an inconsistency  > is noted.  > [snip]M > In the second case (the simple defining of a single clusterwide logical via L > $CRELNM (which DCL DEFINE ultimately uses, of course)], the replication isP > synchronous; the $CRELNM call does not complete until all of the SCS messagingH > needed to deal with replication has completed.  If a node is unable toO > replicate the logical name, it will die on its sword and crash.  The point is Q > to guarantee that the logical name is faithfully replicated across the cluster.   C In such case, then, I should think that Phillip's example should be > forwarded to Engineering for a review. He shows two successiveB DEFINE/<cluster> commands, with no %DCL-I-SUPERSEDE message, which6 should indicate some kind of bug, errr, "feature", no?   --   David J Dachtera dba DJE Systems  http://www.djesys.com/  ) Unofficial OpenVMS Hobbyist Support Page: " http://www.djesys.com/vms/support/  ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/    ------------------------------  $ Date: Sat, 6 Nov 2004 13:48:06 -0500# From: "John Smith" <a@nonymous.com>  Subject: Re: Performance Tuning , Message-ID: <R6idnRciaavzgRDcRVn-sQ@igs.net>   Michael Unger wrote:+ > On 2004-11-06 17:17, "Main, Kerry" wrote:  >  >>' >> "OpenVMS has always had integrity .. ! >> Now, Integrity has OpenVMS .."  > G > That should be "HP OpenVMS has always had integrity. Now HP Integrity  > servers have HP OpenVMS      Why bring HP into it at all?L HP is simply the owner of VMS this week, and that could change soon if rumor is to be believed.   ------------------------------   Date: 6 Nov 2004 17:28:16 -0800 ' From: timasmith@hotmail.com (Tim Smith)  Subject: Re: Performance Tuning < Message-ID: <a7234bb1.0411061728.7bb1ccb@posting.google.com>  t "Alex Daniels" <AlexNoSpamDaniels@themail.co.uk> wrote in message news:<418ce12c$0$27536$db0fefd9@news.zen.co.uk>...B > "Michael Unger" <spam.to.unger@spamgourmet.com> wrote in message' > news:2v44kqF2hapifU1@uni-berlin.de...  > <SNIP>
 > >(There has H > > been a document called "Getting the most out of your Processor" some5 > > time ago; I don't know if it is still available.)  > >  >  > It is still available. > 4 > http://h71000.www7.hp.com/doc/73final/sup/esi.html >  > Alex  5 Thanks for the great articles Alex - I appreciate it!    Tim    ------------------------------  $ Date: Sun, 7 Nov 2004 06:22:57 +02000 From: "Guy Peleg" <guy.peleg@remove_this_hp.com>0 Subject: Re: Request for feedback - COPY utility* Message-ID: <418da510@usenet01.boi.hp.com>  9 "Jim Strehlow" <JimStrehlow@data911.com> wrote in message 6 news:4b6ec350.0411050931.eebd1be@posting.google.com...= > "Guy Peleg" <guy.peleg@remove_this_hp.com> wrote in message & news:<418a0010@usenet01.boi.hp.com>...5 > ...performance improvements for the copy utility...  >  > Can you increase9 >  - the largest block size previously created by $BACKUP I >    Currently I can only $COPY a backup saveset up to 32256 blocks large  >   : We are cooking something in this arena for V.next as well. Stay tuned....  3 >  - the /NETWORK_BLOCK_COUNT to be larger than 127  > A > Let us know what we need to change for the OpenVMS user quotas.  > 0 > Thank you for your efforts to improve OpenVMS. > : > Jim Strehlow, Data911 OpenVMS and Database Administrator > Alameda, CA, USA   ------------------------------  % Date: Sun, 07 Nov 2004 00:32:15 -0600 2 From: David J Dachtera <djesys.nospam@comcast.net>0 Subject: Re: Request for feedback - COPY utility+ Message-ID: <418DC16F.873EBBA0@comcast.net>    Guy Peleg wrote: > ; > "Jim Strehlow" <JimStrehlow@data911.com> wrote in message 8 > news:4b6ec350.0411050931.eebd1be@posting.google.com...? > > "Guy Peleg" <guy.peleg@remove_this_hp.com> wrote in message ( > news:<418a0010@usenet01.boi.hp.com>...7 > > ...performance improvements for the copy utility...  > >  > > Can you increase; > >  - the largest block size previously created by $BACKUP K > >    Currently I can only $COPY a backup saveset up to 32256 blocks large  > >  > < > We are cooking something in this arena for V.next as well. > Stay tuned....   EXCELLENT!!!   --   David J Dachtera dba DJE Systems  http://www.djesys.com/  ) Unofficial OpenVMS Hobbyist Support Page: " http://www.djesys.com/vms/support/  ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/    ------------------------------  * Date: Sat, 6 Nov 2004 18:56:24 +0000 (UTC)P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply)Y Subject: Re: Sys$Startup:License_Check.Exe (revisited) - OPENVMS-ALPHA not authorized on  $ Message-ID: <cmj6oo$3l0$1@online.de>  ; In article <d69b99f3.0411040539.d9735e@posting.google.com>, / kor.rinkens@vodafone.com (Kor Rinkens) writes:    D > Mounting the disks in the sylogicals.com is alright, after this beE > sure to have the lmf$license logical defined in the sylogicals.com. C > If you have this then it should be alright. This works on our vms 
 > clusters  E Right.  Actually, I define it not in SYLOGICALS.COM, but rather in a  * procedure on the newly mounted shadow set.  H > Maybe you have put a include is "node:" in the for example the volshadG > license for each node in the cluster. If you do not do this the first E > node to boot can consume all the units for that particular license.   G This doesn't seem to be happening.  I admit I haven't thought of this,  I but I don't think it is a problem with the hobbyist license.  I only use  4 /INCLUDE for the VAX-VMS and OPENVMS-ALPHA licenses.   ------------------------------   End of INFO-VAX 2004.618 ************************