1 INFO-VAX	Thu, 13 Apr 2006	Volume 2006 : Issue 204       Contents:. Re: Best way to scan processes on remote nodes. Re: Best way to scan processes on remote nodes+ Re: Cluster , locks and lock granting order  Re: DECnet IV routing errors Re: DSPP and OpenVMS media Re: DSPP and OpenVMS media, Re: How does a fix become a published patch?' Re: Internal to External PID conversion ' Re: Internal to External PID conversion ' Re: Internal to External PID conversion ' Re: Internal to External PID conversion E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). E Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64). * Re: manually resetting file revision dates" Re: Shadowed System disk questions& Re: SMTP.CONFIG and SPAM-Action: OPCOM. Re: SMTP: stop sending "no such user" messages$ Re: SoyMail & insufficient privilege$ Re: SoyMail & insufficient privilege Re: VAMP board changes  F ----------------------------------------------------------------------  % Date: Wed, 12 Apr 2006 16:15:13 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> 7 Subject: Re: Best way to scan processes on remote nodes , Message-ID: <443D5FC7.B27B0ECA@teksavvy.com>   shofu_au@yahoo.com.au wrote:D > I have four non-clustered VMS (8.2) systems on which users can runE > process.  I want to limit the users to only running one instance of + > there program on a single node at a time.   E As someone else mentioned, write a command procedure which you invoke  via decnet.      roughly:  test.com =  F $set proc/priv=(noworld,nogroup)    ! make sure you only scan for your own processes  $context= ""
 $found = 0 $loop: $	pid = f$pid(context)# $	if pid .eqs. "" then goto endloop " $	image = f$getjpi(pid,"IMAGNAME")H $	if image .eqs. "$2$dka200:[dir.dir2.dir3]chocolate.exe" then found = 1
 $goto loop $!	 $endloop: I $if found .eq. 1 then write sys$net "FOUND" else write sys$net "NOTFOUND"  $EXIT     # So, when a user logs in, you would     $OPEN/READ temp NODE::"0=TEST" $READ temp buffer  $CLOSE temp G $if buffer .eqs. "FOUND" then write sys$output "You are already running  this on node NODE"    F Now, you would have to run this for every node whenever someone logged$ in. So it would take time to log in.    > Another way to do this wouild be to have 1 node hacve a server= application. Then, whenever someone ran your application, the E application itsefl woudl open a decnet link to that server object and H register itself with the username. When the application crashes or exitsH nicely, the decnmet link is severed and the server knows the application is no longer running.   @ So this way, when someone logs in, you could run a simple clientH application that asks the server process if that user is already runningH that app and the servcer process woudl scan ist memory structures to seeC if there is a decnet link already up for that user and report back.   B This woudl be much quicker because not onlyt would it involved theF login.com establish only 1 decnet link, but it would establish it to aD process that was already running (the server process) so it would be lightning quick.   ------------------------------  # Date: Wed, 12 Apr 2006 21:41:37 GMT , From: Hoff Hoffman <hoff-remove-this@hp.com>7 Subject: Re: Best way to scan processes on remote nodes 2 Message-ID: <lue%f.6237$Oo5.5456@news.cpqcorp.net>   shofu_au@yahoo.com.au wrote:  D > I have four non-clustered VMS (8.2) systems on which users can runE > process.  I want to limit the users to only running one instance of + > there program on a single node at a time.  > G > The user login in via a captive account so I would like to in DCL via G > the login.com that they execute return information to thm saying that < > there are already running their program on the other node. > = > The users sometimes forget they were using node X previous.   G    Users do that.  How are these folks connected in?  DECnet?  Are the  E DECnet or LAT or IP connections not dropping some how?  (That should  ( cause the processes to exit, of course.)  1 > Each program sets a unique process name prefix.   C    There is no such thing as a unique process name, unfortunately.  I Process names are only usually unique within a particular UIC group, but  C the process names are most definitely not unique across UIC groups.   * > What is the best way to achieve this?     I    The best way -- by far -- is to cluster the four nodes.  Then you can  F use direct system service calls, or you can use existing UAF settings C for the user.  But if clustering the nodes is not available or not  5 feasible here, you'll be using some of your own code.   ) > DECnet proxies are setup and available.  > F > I am after a clean way to do this that does not use temporary files.  G    DCL task-to-task connections can provide this, though you'll end up  E raising a connection to each host, and running a server process, and  F retrieving and processing the results, or the inability to connect to G the host.  There are examples of DCL task-to-task around, including in  ; on the Freeware and at the old OpenVMS Ask The Wizard area.   D > My current approach executes a procedure on the remote machine andF > captures the output from show process "name" and read it back in andB > process it.  There must be a better way to do this?  Any lexicalF > function that allow me to scan for processes on non-clustered remote
 > nodes.    :    f$context and f$pid can provide this mechanism for you.  . > Can I use a system service to do this via C?  C    Yes.  sys$process_scan and sys$getjpi[w], or just a wild-carded  G sys$getjpi[w], can provide this.  If you have C code, you can maintain  F connections to a remote server, and basically roll your own directory  server.   G    A little further afield here, you could look at using LDAP for this  A sort of thing, or otherwise maintaining a count in a distributed  G environment.  (Cleaning up after a node failure gets interesting here,  H however -- with a cluster, you can use locks to coordinate and clean up I after a failure.)  Or you could establish a connection into a CGI script  F via a web server on each of the hosts, and look for the user that way.  C    I'd be tempted to look at one of the available terminal timeout  I programs, also known as idle process killers or such, too.  Using one of  F these, you can conceivably even have these connect together to locate F these wandering and forgetful users.  But a generic timeout tool will & eventually clear up stale logins, too.   ------------------------------  % Date: Wed, 12 Apr 2006 17:16:31 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> 4 Subject: Re: Cluster , locks and lock granting order+ Message-ID: <443D6E21.CE91515@teksavvy.com>    Hoff Hoffman wrote: F >    These (apparent) OpenVMS VAX V5.5 nodes have served well over theH > past dozen or more years, and I would think are now worthy of a proper* > and formal and much-deserved retirement.  E It is the DOC that is 5.5, not the nodes. After that, Digital stopped E shipping the "wall" and went to paperback manuals that just aren't as " good as the real documentation...   I >    There are IP routers that provide this exact service, if you need to I > run in those (rather older) OpenVMS configurations that lack current IP  > features and capabilities.  H In order to get such functionality, one has to go to "real" routers, notA consumer level affordable routers. And those do not have the lock 2 manager that instantly notices a node's departure.  F Also, if you have memory limited machines, you might want to optimallyD run the web server on 1 node, the FTP/DNS on another and SMTP on the6 other. And only combine services when one node fails.   	 > (Or you  > can use the load broker   H Load broker is DNS based. Not usable with NAT routers. Also, load brokerE isn't smart enough to shift individual services around. It shifts the G DNS definition of an IP and all services pointing to that IP are moved. D If it is those very services that are consuing the resouces, you areA moving the problem to another node instead or spreading the load.   I >    I'd not tend to target this optimization first.  I'd tend to get the B > configuration working, and would then look for the bottlenecks.   H Oh, I have it working. The problem is that with the cluster alias, I endC up having one iddle system not doing anything, just waiting for the D other to fail. And cluster alias is only 1 IP. I want the ability toG have one IP per service so that I can shift individual services around.   F > these old VAX systems are also very likely going to have 10  megabit	 > NICs.     F Still faster that what the ISP pipes provide, so not a bottleneck yet.E (upstream, Bell Canada limits it to 800kbps, even if downstream it is  between 3 and 5 mbps).    G >    I'd also look at the relative up-time involved, as most any recent G > OpenVMS hardware can operate quite nicely and quite reliably for very  > long durations    B My allmighty Microvax II served me well for 18 years and was stillD working well when I decided to retire it. In fact, it gave me easierG SCSI disk options than the DSSI based systems that replaced it. (mostly : becayuse the DSSI systems lack standard power connectors).    D > in what capabilities are available -- unless the configuration canH > manage the NAT routing, or unless the web server is re-coded to return1 > another address accessible via the remote host    H This isn't so much about disaster tolerance. It is more about having theF capability to shutdown a node for maintenace while seamlessly shifting@ services to another node and have those services return once the maintenance is done.    F >   (I could get the locks to work, but if there's only one IP addressJ > available to route messages, there will be some interesting requirements= > within the TCP/IP Services configuration on the local LAN.)    That is what NAT is for.  G The router gets an incoming WAN call to port 80. Its tables tells it to ( route it to the LAN ip address 10.0.0.15H It gets an incoming WAN call to port 25, its tables tells it to route it to LAN ip address 10.0.0.16   H Whoevere hold ip 10.0.0.16 will accept incoming SMTP calls at that time.> Whoever holds ip address 10.0.0.15 will get HTTP/WEB requests.  F Those would be secondary IPs in hosts, not their primary IPs. So I canG use IFCONFIG to dynamically add/remove them from each host.  Hence, one H host could have its own IP, + 10.0.0.15 + 10.0.0.16 and handle both HTTPG and SMTP requests. Or I can setup 10.0.016 on another node so that SMTP  goes to the other node.     H >   There have been a few enhancements to OpenVMS since V5.5 shipped wayE > back in 1991, and (as I think you've noticed :-) two architectures.   C I just didn't expect long established services like $getjpi to have G lacked just basic stuff as the nodename back in 5.5.  I am aware of NEW A system services that were added since then, and printed their doc H separately. Buit didn't really expect such core systems services to have changed. Now I know.  I >    Yes.  These are the events I was referring to in my response, when I G > indicated there were events that were generated when nodes arrived in 2 > the cluster, and when nodes exited the cluster.   H Since I am not running vintage VMS as you say, what documentation/systemD service must I look at to get such events of nodes joining/leaving ?> (and while I am at it, what is the SYS$ equivalent of F$CSID ?  D The problem with the current documentation format is that one cannotD just sit down and read, quickly flip pages. It is good when you knowF what you are looking for. And that is what I have kept those grey wall books.   ------------------------------    Date: 12 Apr 2006 12:14:52 -0700$ From: "AEF" <spamsink2001@yahoo.com>% Subject: Re: DECnet IV routing errors C Message-ID: <1144869292.134738.129360@u72g2000cwu.googlegroups.com>    Colin Butcher wrote:J > Phase IV sets all adapters known to the NCP database to use the same MACK > address derived from the DECnet address (area*1024 + number, converted to N > hex, byte reversed, used to fill in the xx-xx of AA-00-04-00-xx-xx). Phase VL > will allow you to select which adapters to enable / disable Phase IV styleJ > addressing on (and thus changing the MAC address on those adapters where+ > you've enable Phase IV style addressing).  > E > If you have more than one adapter running with the same MAC address I > connected to the same network segment (even intermittently), you have a D > duplicate MAC address on the LAN segment. That's not a good thing.  F Why do you have multiple machines with the same MAC address (or DECnet address) in the first place?  2 What does routing vs. a bridge have to do with it?  K > When DECnet starts the adapter it is of course smart enough to check that J > no-one else is using the same MAC address already, however, that doesn't6 > always work in these days of "intelligent" switches. > M > So, that might be one way in which a node is seeing it's own hello packets. J > I can't immediately think of a way in which a node with a single adapterI > could see it's own hello packets (or would want to) without the network  > doing something slightly odd.   A So if it sees it own packets, then those packets suddenly stop -- & that's losing adjacency to one's self?   >  > -- >  > Hope this helps, Colin.    We're getting there!  + > colin DOT butcher AT xdelta DOT co DOT uk G > It's not mine, but I like this definition: Legacy = stuff that works. 1 > "AEF" <spamsink2001@yahoo.com> wrote in message ? > news:1144711479.606574.313130@v46g2000cwv.googlegroups.com...    ------------------------------  % Date: Wed, 12 Apr 2006 14:36:30 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> # Subject: Re: DSPP and OpenVMS media , Message-ID: <443D48AD.FC193DD2@teksavvy.com>   Mark Schafer wrote: B > To answer another thread, I have no connection with the hobbyistD > program.  DSPP doesn't allow me to expend resources except to DSPP > company members.  ( That is reasonable. But the question is:  D Does DSPP use its own proprietary download system, or does it use anH existing download service that HP provides for authenticated customers ?   ------------------------------  # Date: Wed, 12 Apr 2006 21:23:23 GMT , From: Hoff Hoffman <hoff-remove-this@hp.com># Subject: Re: DSPP and OpenVMS media 2 Message-ID: <fde%f.6236$9p5.2893@news.cpqcorp.net>   Rich Jordan wrote:H >>   HP has not traditionally provided a consolidated media subscriptionF >> service.  OpenVMS I64 follows this model, and does not have a media >> subscription offering.  >  > Hoff, E >       it is yet another service downgrade/price increase on the VMS ? > customers, and in this case specifically on the VMS developer E > community.  Just like the license transfer fee increase, it hits us H > small timers hardest.  Maybe its just a case of HP continuing to prune; > away at higher level services that it inherited from DEC.   H    (Again) based on what I have been told, the license and the platform I purchase costs for OpenVMS I64 are rather lower than they have been with  G OpenVMS Alpha, and the associated changes are to follow industry norms  I (and HP corporate norms) for software and hardware product licensing and  	 upgrades.   G    Customers with support contracts can and do have access to Internet  F kit downloads, and can also request ("pull") the media as part of the G software support contract.  Conversely, those OpenVMS sites that don't  F have a software support contract or that don't have the (traditional, I and where available) RTU Right-to-Upgrade license or that haven't made a  C software repurchase, don't have a direct way to upgrade to a newer   OpenVMS release.  D    Within intending to give away anything that's not already widely I known about licensing on OpenVMS, the existing LMF software license keys  H have typically been administratively (contractually) tied to an product > release and/or to a support contract, but these PAKs have not H specifically locked out newer product releases that were not authorized @ by the particular associated agreements.  This gets back to the C "management" versus "enforcement" nature of the LMF mechanisms, of  G course.  It is easily possible to lock out new releases within LMF, of  H course, but that means some amount of periodic re-registration work for < the parties involved with a software contract customer site.  H    If you have a software support contract, the new system works pretty G much as it always has, and (with the typical contract information I am  H aware of) a customer does have the rights to the current versions.  The D difference is in the need to "pull" or to download the new releases.   ------------------------------  # Date: Wed, 12 Apr 2006 21:52:53 GMT , From: Hoff Hoffman <hoff-remove-this@hp.com>5 Subject: Re: How does a fix become a published patch? 2 Message-ID: <VEe%f.6240$dq5.2014@news.cpqcorp.net>   Steve Matzura wrote:  F > Where were you in 2001?????????  We know all of this now, of course,7 > as it was explained to us by HP software engineering.   E    Um, in OpenVMS Engineering?  :-)  That employment status aside, I  H wasn't the (as the company was then known) Compaq OpenVMS engineer that C received the particular escalation, and I haven't made more than a  B cursory look at the report other than a very quick look for it in  response to the posting here.   F    I did end up entirely re-implementing the old WRITEBOOT stuff more F recently than 2001, but that's another discussion.  Prior to versions G that are aware of that OpenVMS update, BACKUP effectively includes its  H own embedded version of WRITEBOOT.  On OpenVMS I64, BACKUP now uses the C new (common) SETBOOT pieces, and these SETBOOT pieces are also now  H becoming part of the OpenVMS Alpha releases -- though BACKUP may or may * not be using them there yet, I don't know.   ------------------------------  # Date: Wed, 12 Apr 2006 20:51:03 GMT , From: Hoff Hoffman <hoff-remove-this@hp.com>0 Subject: Re: Internal to External PID conversion2 Message-ID: <XKd%f.6235$bo5.4624@news.cpqcorp.net>   Michael Moroney wrote:0 > Hoff Hoffman <hoff-remove-this@hp.com> writes: >  >> Alan Frisbie wrote:C >>> Is there any way in DCL to convert an Internal Process ID (PID) & >>> to the corresponding External PID?    I    Ugh.  Missed the "in DCL" part.  (My response assumed a kernel-linked  I image was involved.)  Out of curiosity, how did you get into a situation  H where you are in user mode DCL and with an IPID?  IPIDs tend to be from  inner-mode code, obviously.    ------------------------------  # Date: Wed, 12 Apr 2006 21:55:54 GMT , From: Hoff Hoffman <hoff-remove-this@hp.com>0 Subject: Re: Internal to External PID conversion2 Message-ID: <KHe%f.6241$Go5.4568@news.cpqcorp.net>   Alan Frisbie wrote:   < > I'm trying to write a DCL procedure that determines if its> > FT terminal device is a DECwindows terminal or via SSH, etc.? > The PID conversion is the crucial link I am missing -- I have ? > everything else.   I could certainly write the code to do it, 2 > but DCL would be so much more portable and easy.  E    Can you expand on this requirement, with an eye toward specifying  F what you are doing here to make this determination, and how that then I ends up handing you the IPID?  Put another way, there may be a different   way...   ------------------------------   Date: 12 Apr 2006 23:09:50 GMT. From: JONESD@ecr6.ohio-state.edu (David Jones)0 Subject: Re: Internal to External PID conversion: Message-ID: <e1k1bu$mi1$1@charm.magnus.acs.ohio-state.edu>  G In message <KHe%f.6241$Go5.4568@news.cpqcorp.net>, Hoff Hoffman writes:  >Alan Frisbie wrote:= >> I'm trying to write a DCL procedure that determines if its ? >> FT terminal device is a DECwindows terminal or via SSH, etc. @ >> The PID conversion is the crucial link I am missing -- I have@ >> everything else.   I could certainly write the code to do it,3 >> but DCL would be so much more portable and easy.  > E >   Can you expand on this requirement, with an eye toward specifying F >what you are doing here to make this determination, and how that thenI >ends up handing you the IPID?  Put another way, there may be a different  >way...   L I imagine he is getting the IPID of the FT terminal's controlling process byM fetching UCB$L_CPID from that device's UCB (the cell overlays UCB$_LOCKID, so / f$getdvi will return it using the LOCKID code).       < David L. Jones               |      Phone:    (614) 271-6718- Ohio State University        |      Internet: L 140 W. 19th St.              |               jonesd@er6s1.eng.ohio-state.edu: Columbus, OH 43210           |               vman+@osu.edu  1 Disclaimer: I'm looking for marbles all day long.    ------------------------------  % Date: Wed, 12 Apr 2006 18:14:41 -0700 4 From: Alan Frisbie <Usenet02_REMOVE@Flying-Disk.com>0 Subject: Re: Internal to External PID conversion% Message-ID: <1144890876.586391@smirk>    David Jones wrote:I > In message <KHe%f.6241$Go5.4568@news.cpqcorp.net>, Hoff Hoffman writes:  >  >>Alan Frisbie wrote:  >>= >>>I'm trying to write a DCL procedure that determines if its ? >>>FT terminal device is a DECwindows terminal or via SSH, etc. @ >>>The PID conversion is the crucial link I am missing -- I have >>>everything else.   E >>  Can you expand on this requirement, with an eye toward specifying G >>what you are doing here to make this determination, and how that then J >>ends up handing you the IPID?  Put another way, there may be a different >>way...  N > I imagine he is getting the IPID of the FT terminal's controlling process byO > fetching UCB$L_CPID from that device's UCB (the cell overlays UCB$_LOCKID, so 1 > f$getdvi will return it using the LOCKID code).   B That is exactly what I am doing, thanks to the hints you gave hereE a couple of months ago, David.   Thank you very much for those hints.   ? If there is another (supported?) way to determine this, I would = be very happy.   Poking around in a UCB is not what I think I  should be doing at this level.  D Other information I would like to have is which workspace the windowB is in, etc.   When I log in, I want to automatate the setup of allF my windows as much as possible.   If I'm logging in via SSH or Telnet,) I usually want to do something different.    Alan   ------------------------------  % Date: Wed, 12 Apr 2006 14:23:26 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64)., Message-ID: <443D459D.B1C31FB9@teksavvy.com>   Robert Deininger wrote: G > This would be a non-trivial qualification effort for VMS Engineering.   F If you scale up to 32 already, is there really a difference if you add% another 10 CPUs, or another 32 CPUs ?   E I assume that the VMS code itself can already handle 64 CPUs ? (Would / the code itself allow for 256 CPUs ? (1 bytes))   G Would qualification really be for the purpose of testing performance as ' opposed to making sure VMS runs on it ?   Q > I expect qualification, if and when it happens, will be for VMS V8.3, not V8.2.   F If qualification happens, I woudl assume the version would be dictated by the customers.   J > Folks who need M64 systems should definitely be telling their HP accountJ > reps, and if they are running a Cerner application, their Cerner reps as > well.   G This is where there may be a disconnect. Are those sites "HP customers" F or are they "Cerner customers" ? In the later case, they would take toB their Cerner reps and it is Cerner that would need to talk to you.   ------------------------------  % Date: Wed, 12 Apr 2006 15:37:43 -0400 * From: "FredK" <fred.nospam@nospam.dec.com>N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64)., Message-ID: <443d5708$1@usenet01.boi.hp.com>  : "JF Mezei" <jfmezei.spamnot@teksavvy.com> wrote in message& news:443D459D.B1C31FB9@teksavvy.com... > Robert Deininger wrote: I > > This would be a non-trivial qualification effort for VMS Engineering.  > H > If you scale up to 32 already, is there really a difference if you add' > another 10 CPUs, or another 32 CPUs ?  >   A Yes, unless scaling is more-or-less linear.  Typically it is not.   G > I assume that the VMS code itself can already handle 64 CPUs ? (Would 1 > the code itself allow for 256 CPUs ? (1 bytes))  >   K And you would be wrong.  CPUs are also represented as a bitmask.  Expanding D the 32-bit mask to 64 is simple, larger isn't too bad but more work.  I > Would qualification really be for the purpose of testing performance as ) > opposed to making sure VMS runs on it ?  >   K It depends on the definition of "runs".  Would VMS work with 64 CPUs?  Yes. L Will it scale well to 64 CPUs?  That depends on what the workload is and the	 platform.   I > > I expect qualification, if and when it happens, will be for VMS V8.3, 	 not V8.2.  > H > If qualification happens, I woudl assume the version would be dictated > by the customers.  >   L It would likely be dictated by infrastructure changes (for example - the CPUE database and representation) in the OS versions needed to support it.   L > > Folks who need M64 systems should definitely be telling their HP accountL > > reps, and if they are running a Cerner application, their Cerner reps as	 > > well.  > I > This is where there may be a disconnect. Are those sites "HP customers" H > or are they "Cerner customers" ? In the later case, they would take toD > their Cerner reps and it is Cerner that would need to talk to you.  # They are both, and we talk to both.   K The question really is if going to more than 64 CPU's in the same partition J is a large enough win for the application.  The rest then simply becomes a, matter of testing the changes to support it.  G It is highly likelly to see large CPU count support on Itanium - simply L because of trends in multiple-thread/core chips.  Support for the 64-CPUs onH Alpha is likely to be based on having a set of customers with a workload" that scales well on such a system.   ------------------------------  % Date: Wed, 12 Apr 2006 16:27:57 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64)., Message-ID: <443D62C2.695D4688@teksavvy.com>   FredK wrote:M > And you would be wrong.  CPUs are also represented as a bitmask.  Expanding F > the 32-bit mask to 64 is simple, larger isn't too bad but more work.  H Ok, so currently, there are real limited of 32 CPUs coded in the core of VMS ?   F I figured it was like the 96 node cluster "limit" which isn't really aG limit  in the code, it is just a limit in the documentation/support and J some people have been known to cheat on that limit and things worked well.  F So what youa re saying is that VMS currently won't allow you to exceed
 the 32 CPUs ?   M > It depends on the definition of "runs".  Would VMS work with 64 CPUs?  Yes. N > Will it scale well to 64 CPUs?  That depends on what the workload is and the > platform.   C Wasn't the GS series done so it could support 64 CPUs ? Would Tru64 A support 64 CPUs ? Linux ? If one of them actually did have 64 CPU A support, would the scalability/perfornmance of that system be any # indication on how VMS would perform   D (eg: If Tru64 had a 50% performance increase when you increased CPUsH from 32 to 64 (100% increase), would you expect VMS to also have roughly8 a 50% performance increase if CPUs went from 32 to 64 ?   G Or would the performance scalability be *really* totally different from 
 OS to OS ?  N > It would likely be dictated by infrastructure changes (for example - the CPUG > database and representation) in the OS versions needed to support it.   A Theoretically speaking, if you did need to change code to allow a D bitmask of 64 bits, could this be easily done as a patch to existingF versions, or would it require the recompile of so many modules that itJ would be much much simpler to just wait for the next version to come out ?  I > It is highly likelly to see large CPU count support on Itanium - simply N > because of trends in multiple-thread/core chips.  Support for the 64-CPUs onJ > Alpha is likely to be based on having a set of customers with a workload$ > that scales well on such a system.  H Would the code that supports multiple CPUs be platform specific of wouldG it be part of the common source pool (aka: if you change the bitmask to = support 256 CPUs for that IA64 thing, would that code also be - automatically embedded in the Alpha version ?    ------------------------------  % Date: Thu, 13 Apr 2006 06:53:30 +0800 3 From: "Richard Maher" <maher_rj@hotspamnotmail.com> N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64).1 Message-ID: <e1k0bt$cm6$1@news-02.connect.com.au>    Hi Dave,  & Thanks for taking the time to explain.   > H > They support clusters, but they use Oracle for their database and theyF > have problems going to a multiple-active-instance RAC configuration.  H So a) push the envelope and build a new supercomputer b) Continue to get) Oracle and VMS to fix the software. Hmmm?   D Again, I don't want to waste everyone's time with idle curiosity andH frivolous speculation (it's not the ITRC after all) but If someone wouldK like to discuss the problems with getting multiple-active-instances and RAC J (or Grid) working with VMS then I would be very interested. (I'm sure alotL of VMS users not using the Cerner applications would be interested as well!)  I What happened to that disaster tolerant proof-of-concept whitepaper? Only ( one active instance? Where's Kerry Main?   Regards Richard Maher   4 "Dave Harrold" <DHarrold@wi.rr.com> wrote in message2 news:pi6q32p8mopd5cas5inkg165uvuhm7a0l7@4ax.com...5 > On Wed, 12 Apr 2006 17:47:42 +0800, "Richard Maher" & > <maher_rj@hotspamnotmail.com> wrote: >  > >Hi, > > I > >Just out of curiosity, does Cerner not support Clusters or do they run  someJ > >massively-parallel number-crunching stuff that is hitting the wall with 32 > >CPUs? > H > They support clusters, but they use Oracle for their database and theyF > have problems going to a multiple-active-instance RAC configuration. >  > > L > >Does it do any I/O? Own database engine? Is this a patient records system orG > >imaging or key-hole surgery management or mammagram analyzer? Genome 
 > >modelling?  > > K > >What does this beastie do exactly? I'm sure there is a definite need for  a * > >bigger box; just curious about the app. > >  > C > It is an electronic medical record application.  Think manage the F > record and workflow through a medical unit.  E.g.  Scheduling, order
 > entry, etc.  > 7 > We use it to manage our 14 hospitals and 130 clinics.  >  > >Regards Richard Maher > >  >  > 1 > Hope that gives you an idea of the application.  >  > Dave Harrold >  > L ............................................................................ ..7 > David Harrold                                 E-Mail:  David.Harrold@aurora.orgG > Lead Software Systems Engineer                 Phone:           (414)  647-6204G >                                                Pager:           (414)  941-4634G > Aurora Health Care                               Fax:           (414)  647-4999 > 3031 W. Montana Street > Milwaukee, WI 53215    ------------------------------  % Date: Wed, 12 Apr 2006 18:34:50 -0400 * From: "FredK" <fred.nospam@nospam.dec.com>N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64)., Message-ID: <443d808b$1@usenet01.boi.hp.com>  : "JF Mezei" <jfmezei.spamnot@teksavvy.com> wrote in message& news:443D62C2.695D4688@teksavvy.com... > FredK wrote:D > > And you would be wrong.  CPUs are also represented as a bitmask.	 Expanding H > > the 32-bit mask to 64 is simple, larger isn't too bad but more work. > J > Ok, so currently, there are real limited of 32 CPUs coded in the core of > VMS ?  >   I Coded everwhere since SMP was first added.  There are CPU masks in system  service calls as well.  H > I figured it was like the 96 node cluster "limit" which isn't really aI > limit  in the code, it is just a limit in the documentation/support and L > some people have been known to cheat on that limit and things worked well. > H > So what youa re saying is that VMS currently won't allow you to exceed > the 32 CPUs ?  >   J Correct.  You can partition a system with a larger CPU count into multipleG CPUs as long as the CPU ID is 0-31 within the local partition.  But you + cannot have a CPU with a higher ID than 31.   I > > It depends on the definition of "runs".  Would VMS work with 64 CPUs?  Yes.L > > Will it scale well to 64 CPUs?  That depends on what the workload is and the 
 > > platform.  > 9 > Wasn't the GS series done so it could support 64 CPUs ?   G No.  The Marvel GS series was designed as a scaleable architecture that 9 could build systems as small as a single CPU to 256 CPUs.    >Would Tru64C > support 64 CPUs ? Linux ? If one of them actually did have 64 CPU C > support, would the scalability/perfornmance of that system be any % > indication on how VMS would perform  >   K It would give an imperfect indication at best.  Scaling SMP performance has J many dependencies - none the least of which is what the application itselfI does.  If you have a highly parallel application with lots of threads and H very little shared data - things scale well.  But you need to define theI workload.  The OS itself has an impact as well - how it does locking, how " fine the lock granularity is, etc.  F > (eg: If Tru64 had a 50% performance increase when you increased CPUsJ > from 32 to 64 (100% increase), would you expect VMS to also have roughly9 > a 50% performance increase if CPUs went from 32 to 64 ?  >   J There is no way to give you that simple answer.  If the application ran onK both VMS and Tru64 with 32 CPUs at identical performance - and Tru64 scaled G 50% - then the answer is probably yes.  But those are a lot of if's and  assumptions.  I > Or would the performance scalability be *really* totally different from  > OS to OS ? >   G It depends on what the application(s) are doing.  Each OS tends to have  different strengths.    L > > It would likely be dictated by infrastructure changes (for example - the CPU I > > database and representation) in the OS versions needed to support it.  > C > Theoretically speaking, if you did need to change code to allow a F > bitmask of 64 bits, could this be easily done as a patch to existingH > versions, or would it require the recompile of so many modules that itL > would be much much simpler to just wait for the next version to come out ? >   L There are user interfaces as well (system service calls) that need to changeE as well as utilities like SHOW CPU.  It is also easier to qual in the  context of a release process.   K > > It is highly likelly to see large CPU count support on Itanium - simply E > > because of trends in multiple-thread/core chips.  Support for the 
 64-CPUs onL > > Alpha is likely to be based on having a set of customers with a workload& > > that scales well on such a system. > J > Would the code that supports multiple CPUs be platform specific of wouldI > it be part of the common source pool (aka: if you change the bitmask to ? > support 256 CPUs for that IA64 thing, would that code also be / > automatically embedded in the Alpha version ?   K Yes/no.  There are some low-level differences in the Itanium since there is J firmware software handoff differences, and other things.  In any case, the@ general code would probably be common to maintain user interfaceH compatibility.  That doesn't mean that VMS would automatically work on a 64-way Marvel.   ------------------------------  % Date: Wed, 12 Apr 2006 20:21:28 -0500 @ From: "David J. Dachtera" <djesys.nospam@NeOaSrPtAhMlNiOnWk.net>N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64).6 Message-ID: <443DA798.4BC69D73@NeOaSrPtAhMlNiOnWk.net>   Dave Froble wrote: > [snip]2 > The concept of "If you build it, they will come" > = > has a converse of "If you don't build it, they won't come".   E More like, "If you don't build it, your competition wins since that's  where your customers will go".  F ...on the other hand, it appears that this is exactly what HP intends;1 so, obviously, their plan is working as intended.    --   David J Dachtera dba DJE Systems  http://www.djesys.com/  & Unofficial OpenVMS Marketing Home Page! http://www.djesys.com/vms/market/   ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/   " Unofficial OpenVMS-IA32 Home Page: http://www.djesys.com/vms/ia32/   ) Unofficial OpenVMS Hobbyist Support Page: " http://www.djesys.com/vms/support/   ------------------------------  % Date: Wed, 12 Apr 2006 20:30:09 -0500 @ From: "David J. Dachtera" <djesys.nospam@NeOaSrPtAhMlNiOnWk.net>N Subject: Re: Is OpenVMS certified yet for 64-way Hard partitions (GS1280 M64).6 Message-ID: <443DA9A1.3B1A4149@NeOaSrPtAhMlNiOnWk.net>   Dave Froble wrote: >  > JF Mezei wrote:  > > Dave Froble wrote: > > E > >>If HP cared about keeping these customers, they'd be pro-actively I > >>talking to them and insuring them they'd have whatever they need.  If N > >>that's 64 CPU Alphas until Cerner is running on the itanic, then so be it. > >  > > J > > Consider Swift. Once Swift had agreed to heed's Palmer suggestion theyI > > abandon VMS, there was no turning back, and the owner of VMS wouldn't L > > have lifted a finger to try to retain customers it knew wouldn't be ableI > > to remain on VMS for long anyways. (and during the Compaq era, Compaq I > > would have known the odds were very slim that a Swift customer on VMS I > > would migrate to a windows based solution which didn't fully exist at 4 > > that time, and that is all Compaq had to offer). > > I > > **IF** Cerner doesn't have intentions of remaining on VMS in the long E > > term and prefers to focus on other platforms such as AIX, then HP F > > wouldn't have much in terms of incentives to spend money to retain, > > customers it knows it will lose anyways. > > J > > On the other hand, if the VMS engineers know that VMS works on 64 CPUsF > > but that they just never got around to documenting it, adding thisK > > qualified support might be easy, or they may simply tell Cerner that it @ > > should work and gice Cerner the tools to test it themselves. > I > Jumping to conclusions without any reason.  I've not read one word that F > says Cerner is abandoning VMS.  Not saying they are, or aren't, justI > haven't seen anything to indicate they are.  Unless you have some solid J > information, speculating on one of your pet theories isn't very helpful.  E Well, consider that two of Cerner's largest sites in the Chicago area ; have been lost by HP as storage customers in favor of EMC.    B Consider that Cerner's three largest sites in the Chicago area areF contemplating an eventualy move to AIX to obtain 64 CPU support (and a4 somewhat less uncertain future for their platforms).  G Consider also that we're on the threshold of V8.3 and Cerner has yet to D certify V8.2 (which will be two versions back, allowing for V8.2-1).  F Consider that the last sale dates for Alpha are just around the corner" and Cerner has yet to certify I64.  H I know you're an intelligent guy, Dave. I'm sure you can put 2.0 and 2.0A together come with something approximating 4.0, even allowing for  floating-point errors.  H As someone in my office put it, the "marriage" between Cerner and VMS is seriously on the rocks.    --   David J Dachtera dba DJE Systems  http://www.djesys.com/  & Unofficial OpenVMS Marketing Home Page! http://www.djesys.com/vms/market/   ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/   " Unofficial OpenVMS-IA32 Home Page: http://www.djesys.com/vms/ia32/   ) Unofficial OpenVMS Hobbyist Support Page: " http://www.djesys.com/vms/support/   ------------------------------  # Date: Thu, 13 Apr 2006 01:08:54 GMT , From: Hoff Hoffman <hoff-remove-this@hp.com>3 Subject: Re: manually resetting file revision dates 2 Message-ID: <Gwh%f.6264$cv5.2022@news.cpqcorp.net>   mabbuttg@yahoo.ca wrote:C > I'm testing a script of mine which is checking some file revision I > dates.  For proper testing, I need to create files with a few different  > dates. > H > I did some searching through past newsgroup posts, and found some infoD > on changing the rev. date with some of the freeware apps (DFU) andG > someone posted a C program.  However, we're running a stock 7.3-1/AXP E > install - no freeware, just COBOL installed as an extra (other than  > TCP/IP, of course).  > E > Is there any way to do this on my setup without changing the system 	 > time...   Y http://h18000.www1.hp.com/support/asktima/operating_systems/009189D7-74D47620-1C0069.html Y http://h18000.www1.hp.com/support/asktima/operating_systems/00941061-F1C3F300-1C03C5.html   E    Given your situation, I'd get a test system, and try your testing  G there.  That keeps the auditors happy, and it allows you easier access  F to tools like DFU, and it's also where you can get some compilers and G tools loaded.  And it avoids all the fun that arises when the auditors  F hear you are changing file dates.  But failing that availability, the A above examples are from the support database, and are written in  % Macro32, and you have access to that.    ------------------------------  + Date: Wed, 12 Apr 2006 18:51:59 +0000 (UTC) P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply)+ Subject: Re: Shadowed System disk questions $ Message-ID: <e1ji8e$m27$1@online.de>  F In article <1d45e$443c4868$50db5015$23147@news.hispeed.ch>, Paul Sture' <paul.sture.nospam@hispeed.ch> writes:    F > > Note that you cannot dismount one member of a shadow set on a live* > > system and expect to get a clean copy. > K > Then why have you gone to the trouble to move SYSUAF etc off your system   > disk?   > Mainly so that I have to maintain just one set of these files.  = > Those files at the very minimum will be safe. What else of  K > importance is left? Obviously you need to know your own system to answer  
 >   this one.   G Right.  You might be right that any other open files will probably not  F be opened for write, and that things might be OK.  But it still isn't  officially supported.   K > You can also close applications by running manually closing batch queues  H > and running SYSHUTDWN.COM to ensure they aren't using the system disk.  E True, but if I have to shut down everything, it's not that much more  4 trouble to completely shut down the node and reboot.   > >  You need to shut down theJ > > system, physically disconnect a member, and bring it back up with one  > > less member. > A > No way, except under exceptional circumstances. Rebooting with  I > SHADOW_SYS_DISK=0 suffices. Why much with hardware when you don't need  5 > to? Anyway, the disk farm could be many miles away.   I That will work, of course, if you set SHADOW_SYS_DISK=0 first.  However,  G it's not a dynamic parameter, so a reboot is needed.  Also, one has to  = be careful not to mount the backup member back in by mistake.    ------------------------------  % Date: Wed, 12 Apr 2006 16:18:13 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> / Subject: Re: SMTP.CONFIG and SPAM-Action: OPCOM , Message-ID: <443D607A.91BF8DCE@teksavvy.com>  / Phillip Helbig---remove CLOTHES to reply wrote:  >    SPAM-Action: OPCOM   D > The only parameter which should be rejecting spam at the moment is> > Symbiont-Checks-Deliverability: which I have set to FALSE.    H The SPAM action would be a message blocked because of RBLs, the sender'sE IP not being unbacktranslatable and the sender's domain being forwrad  translatable.   F The deliverability is not a "spam" feature. Consider a legitimate userF sending you a message and maming a typing error. This is not spam. TheH receiver will refuse receipt of message, so the sending SMTP willa dvise( the sender that the username is invalid.   ------------------------------  % Date: Wed, 12 Apr 2006 14:34:23 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> 7 Subject: Re: SMTP: stop sending "no such user" messages , Message-ID: <443D482E.680763AA@teksavvy.com>   david20@alpha2.mdx.ac.uk wrote: = > Not really. Just the difference between content and syntax. L > The syntax of the HELO argument must be correct since the argument is thenC > used in received-lines which expect it to be in a certain format. M > However the value of an argument which is syntacticly correct should not be  > used to reject mail.    H Lets not kid ourselves here. You may not be rejecting the transaction at5 the HELO stage, but you'll be rejecting it elsewhere.   D I thik the point of the RFC telling you that you cannot reject it is= that there are many cases where the HELO data is meaningless.   G For instance, a PC sending a message to the LAN's SMTP server will have  some name like "MyPC".  F Similarly, someone travelling with a  laptop will use a dialup serviceE and legitimately use your SMTP server (possibly with authentication). F The laptop may identify itself as "mylaptop", and this will definitelyJ not match the IP address and owner of the IP address currently being used.    A BUT: if someone pretends to be yahoo.com in the HELO, and you use G Yahoo's technique of checking yahoo'.s DNS entries to get a list of IPs G that are allowed to send on behalf of yahoo, and you find that it isn't C a valid IP, then you'll know this is a forged message and should be  rejected later on.   ------------------------------  + Date: Wed, 12 Apr 2006 17:56:17 +0000 (UTC) < From: gartmann@nonsense.immunbio.mpg.de (Christoph Gartmann)- Subject: Re: SoyMail & insufficient privilege ) Message-ID: <e1jf01$qr5$1@news.BelWue.DE>   ] In article <123qda2bdniu9e4@corp.supernews.com>, Mark Daniel <mark.daniel@vsm.com.au> writes: 
 >You could > ) >   pass /soymail/-/* /www_root/soymail/* ! >   redirect /soymail/ /soymail/~ % >   map /soymail/~* /htbin/soymail/~* 5 >   protect /htbin/soymail/~* www_system:soymail.prot  > G >which would automatically place the tilde into the user's URL (obvious G >to them) without the need for them to do it manually.  This would only G >occur on the startup access (https://the.host.name/soymail/). It would E >then map it appropriately to the /htbin/ path internally keeping you ! >desired path to the application.   M Ok, this is what I currently do but I would prefer something in SOYMAIL.CONF.   G > >In addition, the logout button of soyMAIL isn't working. Users get a C > >message that it is not available on this platform. But I put the H > >"[logout-realm]" in the SOYMAIL.CONF file. Anything else to have this > >working?  > : >The Install and Admin document discusses this issue underC >[logout-realm] in section 3.3.  As long as it looks something like  >  >   [logout-realm]  VMS account  > E >the button should not be italicised and should provide a JavaScript   >alert when activated.  J The button looks pretty normal, but produces the above error when clicked.  / >The 'VMS account' being the realm description  ! >from the soymail.prot file rule.   C To me both strings look identical. Here is what is in SOYMAIL.CONF:   ) [logout-realm]  Enter Username & Password  #    And here is SOYMAIL.PROT:   ! <realm> Enter Username & Password  *@*.*.*.*:443 *    So what is wrong?    Regards,    Christoph Gartmann    --  E  Max-Planck-Institut fuer      Phone   : +49-761-5108-464   Fax: -452   ImmunbiologieI  Postfach 1169                 Internet: gartmann@immunbio dot mpg dot de   D-79011  Freiburg, Germany 9                http://www.immunbio.mpg.de/home/menue.html    ------------------------------  % Date: Thu, 13 Apr 2006 02:36:26 +0930 * From: Mark Daniel <mark.daniel@vsm.com.au>- Subject: Re: SoyMail & insufficient privilege 0 Message-ID: <123qgcijn0jupf1@corp.supernews.com>   Christoph Gartmann wrote: _ > In article <123qda2bdniu9e4@corp.supernews.com>, Mark Daniel <mark.daniel@vsm.com.au> writes:  >  >>You could  >>) >>  pass /soymail/-/* /www_root/soymail/* ! >>  redirect /soymail/ /soymail/~ % >>  map /soymail/~* /htbin/soymail/~* 5 >>  protect /htbin/soymail/~* www_system:soymail.prot  >>H >>which would automatically place the tilde into the user's URL (obviousH >>to them) without the need for them to do it manually.  This would onlyH >>occur on the startup access (https://the.host.name/soymail/). It wouldF >>then map it appropriately to the /htbin/ path internally keeping you" >>desired path to the application. >  > O > Ok, this is what I currently do but I would prefer something in SOYMAIL.CONF.   & Where you bin the last ten weeks?  ;-)  G >>>In addition, the logout button of soyMAIL isn't working. Users get a C >>>message that it is not available on this platform. But I put the H >>>"[logout-realm]" in the SOYMAIL.CONF file. Anything else to have this >>>working?  >>; >>The Install and Admin document discusses this issue under D >>[logout-realm] in section 3.3.  As long as it looks something like >> >>  [logout-realm]  VMS account  >>F >>the button should not be italicised and should provide a JavaScript  >>alert when activated.  >  > L > The button looks pretty normal, but produces the above error when clicked. >  > 0 >>The 'VMS account' being the realm description  > " >>from the soymail.prot file rule. > E > To me both strings look identical. Here is what is in SOYMAIL.CONF:  > + > [logout-realm]  Enter Username & Password  > #  >  > And here is SOYMAIL.PROT:  > # > <realm> Enter Username & Password  > *@*.*.*.*:443 *  >  > So what is wrong?  > 
 > Regards, >    Christoph Gartmann  >    ------------------------------  % Date: Wed, 12 Apr 2006 15:52:32 -0400 3 From: "Richard B. Gilbert" <rgilbert88@comcast.net>  Subject: Re: VAMP board changes : Message-ID: <F4qdnSXljrwcx6DZnZ2dnUVZ_tednZ2d@comcast.com>   Ken Robinson wrote: 2 > On 4/12/06, issinoho <issinoho@gmail.com> wrote: > G >>Ken, I'm having some trouble accessing the site with IE7 today, IE6 &  >>Firefox both look fine.  >  > E > I'm trying with Firefox. Now I'm getting a completely blank screen.  >  > Ken   G My ancient Netscape 7.2 browser worked like a champ a few minutes ago.  H I know it's not politcally correct but I've grown accustomed to it. . .  .   (Like VMS itself)    ------------------------------   End of INFO-VAX 2006.204 ************************