INFO-VAX Wed, 12 Mar 2008 Volume 2008 : Issue 143 Contents: Re: "file locked by another user" mystery Re: "file locked by another user" mystery Re: "file locked by another user" mystery Re: "file locked by another user" mystery Re: "file locked by another user" mystery Re: "file locked by another user" mystery Re: Apache performance problem Re: Apache performance problem Re: Apache performance problem Re: Apache performance problem Re: Apache performance problem Re: Apache performance problem Cheap Gigabit cards at Island Computers Re: DHCP Server question/problem Re: from TheDailyWTF Re: OpenVMS Oracle v7.3.3 to Oracle 10GR2 migration and application development Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Re: Proof that macintosh is better than VMS Time and TDF after DST advance Re: Time and TDF after DST advance Re: Time and TDF after DST advance Re: Time and TDF after DST advance Re: Time and TDF after DST advance Re: Time and TDF after DST advance Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot Re: Time changing after reboot ---------------------------------------------------------------------- Date: Tue, 11 Mar 2008 14:04:27 -0400 From: "Richard B. Gilbert" Subject: Re: "file locked by another user" mystery Message-ID: <47D6C9AB.5090906@comcast.net> tadamsmar wrote: > On Mar 11, 2:49 pm, bri...@encompasserve.org wrote: > >>In article <6360eeaa-a66f-4073-af48-73abb837c...@60g2000hsy.googlegroups.com>, tadamsmar writes: >> >> >> >> >> >> >>>On Mar 11, 11:04=A0am, Volker Halle wrote: >>> >>>>If these are actually shared images, you need to de-install them with >>>>INSTALL DELETE filename >>> >>>>If these are global sections, you need to run the application image, >>>>which deletes those global sections and you need to stop all process, >>>>which still have them mapped. >>> >>>>I documented an analytic approach using SDA to find out, which >>>>processes are still mapped to delete-pending global sections. >>> >>>>Please see this article in ITRC: >>> >>>>http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId... >>> >>>>Volker. >>> >>>Thanks, I will try the ideas in that article the next time it >>>happens. Maybe I can improve my app shutdown command procedure to >>>prevent the problem. >> >>As I suggested earlier, why not try those ideas _now_? >> >>If you wait until the problem occurs, you'll be under the gun and >>might very well end up rebooting the machine due to time constraints. >> >>If you try it now, you'll be able to go slow and work out any >>difficulties with the procedure ahead of time. You'll have had practice. >> >>As a bonus, you can figure out which of your processes do or do not >>have that section file mapped while the system is running in production. >>The ones that do will be your obvious targets when the problem recurs.- Hide quoted text - >> >>- Show quoted text - > > > Good idea, I can certainly try to tighten it up. > > The problem is transient. I don't know how to reproduce it > reliably. It does not happen everytime I shutdown the app. In > fact, it never happened for 20 years, only started happening recently. Do you have any sort of a "change control" process? If so, looking at what changed during the weeks or months before the problem appeared, may provide a clue. If you don't have a change control process, now you have an inkling as to why people do have change control. ------------------------------ Date: Tue, 11 Mar 2008 12:15:38 -0700 (PDT) From: tadamsmar Subject: Re: "file locked by another user" mystery Message-ID: On Mar 11, 2:04=A0pm, "Richard B. Gilbert" wrote: > tadamsmar wrote: > > On Mar 11, 2:49 pm, bri...@encompasserve.org wrote: > > >>In article <6360eeaa-a66f-4073-af48-73abb837c...@60g2000hsy.googlegroups= .com>, tadamsmar writes: > > >>>On Mar 11, 11:04=3DA0am, Volker Halle wrote:= > > >>>>If these are actually shared images, you need to de-install them with > >>>>INSTALL DELETE filename > > >>>>If these are global sections, you need to run the application image, > >>>>which deletes those global sections and you need to stop all process, > >>>>which still have them mapped. > > >>>>I documented an analytic approach using SDA to find out, which > >>>>processes are still mapped to delete-pending global sections. > > >>>>Please see this article in ITRC: > > >>>>http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId.= .. > > >>>>Volker. > > >>>Thanks, =A0I will try the ideas in that article the next time it > >>>happens. =A0Maybe I can improve my app shutdown command procedure to > >>>prevent the problem. > > >>As I suggested earlier, why not try those ideas _now_? > > >>If you wait until the problem occurs, you'll be under the gun and > >>might very well end up rebooting the machine due to time constraints. > > >>If you try it now, you'll be able to go slow and work out any > >>difficulties with the procedure ahead of time. =A0You'll have had practi= ce. > > >>As a bonus, you can figure out which of your processes do or do not > >>have that section file mapped while the system is running in production.= > >>The ones that do will be your obvious targets when the problem recurs.- = Hide quoted text - > > >>- Show quoted text - > > > Good idea, I can certainly try to tighten it up. > > > The problem is transient. =A0I don't know how to reproduce it > > reliably. =A0 It does not happen everytime I shutdown the app. =A0 In > > fact, it never happened for 20 years, only started happening recently. > > Do you have any sort of a "change control" process? =A0If so, looking at > what changed during the weeks or months before the problem appeared, may > provide a clue. =A0If you don't have a change control process, now you > have an inkling as to why people do have change control.- Hide quoted text= - > > - Show quoted text - Yes, I have change control, CMS to be exact. The shutdown code has been very stable. The only thing I can think of that has changed is the size of the global sections. They get bigger when I have to add fields to the real time data base. And I have patched VMS. ------------------------------ Date: Tue, 11 Mar 2008 17:12:18 -0400 From: JF Mezei Subject: Re: "file locked by another user" mystery Message-ID: <47d6f643$0$31259$c3e8da3@news.astraweb.com> Volker Halle wrote: > INSTALL REMOVE (or DELETE) will mark the global section for deletion. > The file will not be closed, until the last process unmaps the global > section. Does INSTALL LIST actually show global section files ? I know that INSTALL LIST/GLOBAL shows global sections. If a program creates a permanent global section mapped to a file, does this automatically show up in INSTALL LIST ? To the original poster: I would look at your application to see why it isn't causing that file to be "freed" from being a global section. You may have to write a small utility to delete the global section (but not the file) in order to "free" the file once your other processes that access the global section are done with it. ------------------------------ Date: 11 Mar 2008 17:22:47 -0600 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: "file locked by another user" mystery Message-ID: In article <487660d5-d06d-4c80-baa7-7be34618bfd0@e23g2000prf.googlegroups.com>, Volker Halle writes: > On 11 Mrz., 20:24, koeh...@eisner.nospam.encompasserve.org (Bob > Koehler) wrote: >> =A0 =A0Then the file is INSTALLed. =A0A privileged user INSTALLed it and >> =A0 =A0a privileged user can INSTALL REMOVE it. =A0At which point it will >> =A0 =A0get closed. > > INSTALL REMOVE (or DELETE) will mark the global section for deletion. > The file will not be closed, until the last process unmaps the global > section. Which is why I said "will get closed" (future tense). And if such a process exists, then the show file command will still show the file as open. SDA can be used to find out what process has the section mapped, but I'll not go into that here. ------------------------------ Date: Tue, 11 Mar 2008 21:48:45 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: "file locked by another user" mystery Message-ID: tadamsmar writes: >On Mar 11, 8:35=A0am, "Ken Robinson" wrote: >> Did you, at least, cause a crash dump, so it could be analyzed? >I did not. >What could I learn from that? So that perhaps one of the many people in here (including myself) who know their away around with SDA could look at the dump and tell you once and for all what the problem was (and come up with a solution). Pretty much any time you have to force a reboot to clear up a mystery problem like this you should save the dump. Better yet, of course, is to get someone to look at the live system and unwedge it without forcing a reboot. Remember, VMS isn't Windows. ------------------------------ Date: Tue, 11 Mar 2008 17:31:36 -0700 (PDT) From: AEF Subject: Re: "file locked by another user" mystery Message-ID: <6b4214f7-2461-4d40-a4fc-ebc5d619dd26@m44g2000hsc.googlegroups.com> On Mar 11, 10:20 am, Jan-Erik S=F6derholm wrote: > tadamsmar wrote: > > On Mar 11, 9:09 am, Jan-Erik S=F6derholm > > wrote: > >> I think you have to provide hard evidence that things realy are > >> like you're saying. > > > Thanks! You helped me solve this problem. > > Fine, I'm gald you did. > > > > > I'll just assume I am wrong! > > I couldn't care less what *you* assume. > But *I* will assume that until you prove your're not. > > Not by what you say/write/claim, but by presenting hard > evidence (error messages, output from commands and so on). > > It's not enough to say that "I did this and that", *SHOW* > that you actualy did it ! Seconded! To the OP: Please show us some command output of what you did. Copy and paste it if you can. AEF > > Anyway, It seems as you've found that the images/sections > are still INSTALL'ed, so that's it, right ? :-) > > Best Regards, > Jan-Erik. ------------------------------ Date: Tue, 11 Mar 2008 13:28:00 -0600 From: John Nebel Subject: Re: Apache performance problem Message-ID: <47D6DD40.40501@csdco.com> Alan Winston - SSRL Central Computing wrote: > In article <47D045E6.7050303@csdco.com>, John Nebel writes: >> This is on OpenVMS 7.3-2 and Apache 1.3-1 >> >> Apache is performing an insane number of logical name translations, this can >> exceed a rate of 200k/second and averages around 10k/second in our environment. >> >> One thing Apache appears to be doing is looping on a translation of >> apache$dcl_sync_xxxxxxxx >> >> 6-MAR 11:55:27.182282 03 4DF1DF06 APACHE_HTTPD 000FA418 APACHE$DCL_SYNC_4DFAA98B >> 6-MAR 11:55:27.182266 03 4DF1DF06 APACHE_HTTPD 000FA418 APACHE$DCL_SYNC_4DFAA98B >> 6-MAR 11:55:27.182250 03 4DF1DF06 APACHE_HTTPD 000FA418 APACHE$DCL_SYNC_4DFAA98B >> 6-MAR 11:55:27.182235 03 4DF1DF06 APACHE_HTTPD 000FA418 APACHE$DCL_SYNC_4DFAA98B >> >> Ultimately this appears to have the ability to jam VMS enough to crash it with a >> cluexit. >> >> Is their a workaround, or does Apache 2.1-1 have a more reasonable behavior? > > > There's a CRTL flag that'll make Apache cache those logical-name translation > results. > > $ define/nolog decc$enable_getenv_cache enable > > in the APACHE$WWW LOGIN.COM > > -- Alan Alan, I don't think the decc$enable_getenv_cache logical will help in this case - the offending code within Apache (proc.c) starts with: ** Loop here until the child creates the synchronizing logical name or until ** the child process has exited. */ while (TRUE) { /* ** Check to see if the child process has created the synchronizing logical ** name; If it has, then exit the loop. */ status = SYS$TRNLNM (&SyncLnmAttr, &SyncTblDesc, &SyncLnmDesc, 0, 0); if (status & 1) break; [cut] The sync logical is defined within apache$dcl.com and the above routine just runs in a tight loop until the definition takes place. At times this can take 30% of a horrendous number of logical name translations. John ------------------------------ Date: Tue, 11 Mar 2008 16:13:14 -0400 From: "Richard B. Gilbert" Subject: Re: Apache performance problem Message-ID: <47D6E7DA.1060107@comcast.net> John Nebel wrote: > > > Alan Winston - SSRL Central Computing wrote: > >> In article <47D045E6.7050303@csdco.com>, John Nebel >> writes: >> >>> This is on OpenVMS 7.3-2 and Apache 1.3-1 >>> >>> Apache is performing an insane number of logical name translations, >>> this can exceed a rate of 200k/second and averages around 10k/second >>> in our environment. >>> >>> One thing Apache appears to be doing is looping on a translation of >>> apache$dcl_sync_xxxxxxxx >>> >>> 6-MAR 11:55:27.182282 03 4DF1DF06 APACHE_HTTPD 000FA418 >>> APACHE$DCL_SYNC_4DFAA98B >>> 6-MAR 11:55:27.182266 03 4DF1DF06 APACHE_HTTPD 000FA418 >>> APACHE$DCL_SYNC_4DFAA98B >>> 6-MAR 11:55:27.182250 03 4DF1DF06 APACHE_HTTPD 000FA418 >>> APACHE$DCL_SYNC_4DFAA98B >>> 6-MAR 11:55:27.182235 03 4DF1DF06 APACHE_HTTPD 000FA418 >>> APACHE$DCL_SYNC_4DFAA98B >>> >>> Ultimately this appears to have the ability to jam VMS enough to >>> crash it with a cluexit. >>> >>> Is their a workaround, or does Apache 2.1-1 have a more reasonable >>> behavior? >> >> >> >> There's a CRTL flag that'll make Apache cache those logical-name >> translation >> results. >> $ define/nolog decc$enable_getenv_cache enable >> >> in the APACHE$WWW LOGIN.COM >> >> -- Alan > > > Alan, > > I don't think the decc$enable_getenv_cache logical will help in this > case - the offending code within Apache (proc.c) starts with: > > ** Loop here until the child creates the synchronizing logical name or > until > ** the child process has exited. > */ > while (TRUE) > { > /* > ** Check to see if the child process has created the synchronizing > logical > ** name; If it has, then exit the loop. > */ > status = SYS$TRNLNM (&SyncLnmAttr, > &SyncTblDesc, > &SyncLnmDesc, > 0, 0); > if (status & 1) > break; > [cut] > > The sync logical is defined within apache$dcl.com and the above routine > just runs in a tight loop until the definition takes place. > > At times this can take 30% of a horrendous number of logical name > translations. > > John Is there some reason why this code can't have a built in wait? Waiting for two or five seconds between checks would save one hell of a lot of CPU that could probably be better used elsewhere. ------------------------------ Date: 11 Mar 2008 20:23:44 GMT From: VAXman- @SendSpamHere.ORG Subject: Re: Apache performance problem Message-ID: <47d6ea50$0$25052$607ed4bc@cv.net> In article <47D6E7DA.1060107@comcast.net>, "Richard B. Gilbert" writes: >John Nebel wrote: >> >> >> Alan Winston - SSRL Central Computing wrote: >> >>> In article <47D045E6.7050303@csdco.com>, John Nebel >>> writes: >>> >>>> This is on OpenVMS 7.3-2 and Apache 1.3-1 >>>> >>>> Apache is performing an insane number of logical name translations, >>>> this can exceed a rate of 200k/second and averages around 10k/second >>>> in our environment. >>>> >>>> One thing Apache appears to be doing is looping on a translation of >>>> apache$dcl_sync_xxxxxxxx >>>> >>>> 6-MAR 11:55:27.182282 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> 6-MAR 11:55:27.182266 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> 6-MAR 11:55:27.182250 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> 6-MAR 11:55:27.182235 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> >>>> Ultimately this appears to have the ability to jam VMS enough to >>>> crash it with a cluexit. >>>> >>>> Is their a workaround, or does Apache 2.1-1 have a more reasonable >>>> behavior? >>> >>> >>> >>> There's a CRTL flag that'll make Apache cache those logical-name >>> translation >>> results. >>> $ define/nolog decc$enable_getenv_cache enable >>> >>> in the APACHE$WWW LOGIN.COM >>> >>> -- Alan >> >> >> Alan, >> >> I don't think the decc$enable_getenv_cache logical will help in this >> case - the offending code within Apache (proc.c) starts with: >> >> ** Loop here until the child creates the synchronizing logical name or >> until >> ** the child process has exited. >> */ >> while (TRUE) >> { >> /* >> ** Check to see if the child process has created the synchronizing >> logical >> ** name; If it has, then exit the loop. >> */ >> status = SYS$TRNLNM (&SyncLnmAttr, >> &SyncTblDesc, >> &SyncLnmDesc, >> 0, 0); >> if (status & 1) >> break; >> [cut] >> >> The sync logical is defined within apache$dcl.com and the above routine >> just runs in a tight loop until the definition takes place. >> >> At times this can take 30% of a horrendous number of logical name >> translations. >> >> John > >Is there some reason why this code can't have a built in wait? Waiting >for two or five seconds between checks would save one hell of a lot of >CPU that could probably be better used elsewhere. This is FN ugly! Is there some reason _why_ this has been implemented in this fashion? There are better ways to inform the parent process that a subprocess has been created instead of looping on a semaphore implemented with a logical name. > -- VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)COM "Well my son, life is like a beanstalk, isn't it?" http://tmesis.com/drat.html ------------------------------ Date: Tue, 11 Mar 2008 14:35:10 -0600 From: John Nebel Subject: Re: Apache performance problem Message-ID: <47D6ECFE.3070309@csdco.com> Richard B. Gilbert wrote: > John Nebel wrote: >> >> >> Alan Winston - SSRL Central Computing wrote: >> >>> In article <47D045E6.7050303@csdco.com>, John Nebel >>> writes: >>> >>>> This is on OpenVMS 7.3-2 and Apache 1.3-1 >>>> >>>> Apache is performing an insane number of logical name translations, >>>> this can exceed a rate of 200k/second and averages around 10k/second >>>> in our environment. >>>> >>>> One thing Apache appears to be doing is looping on a translation of >>>> apache$dcl_sync_xxxxxxxx >>>> >>>> 6-MAR 11:55:27.182282 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> 6-MAR 11:55:27.182266 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> 6-MAR 11:55:27.182250 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> 6-MAR 11:55:27.182235 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>> APACHE$DCL_SYNC_4DFAA98B >>>> >>>> Ultimately this appears to have the ability to jam VMS enough to >>>> crash it with a cluexit. >>>> >>>> Is their a workaround, or does Apache 2.1-1 have a more reasonable >>>> behavior? >>> >>> >>> >>> There's a CRTL flag that'll make Apache cache those logical-name >>> translation >>> results. $ define/nolog decc$enable_getenv_cache enable >>> >>> in the APACHE$WWW LOGIN.COM >>> >>> -- Alan >> >> >> Alan, >> >> I don't think the decc$enable_getenv_cache logical will help in this >> case - the offending code within Apache (proc.c) starts with: >> >> ** Loop here until the child creates the synchronizing logical name or >> until >> ** the child process has exited. >> */ >> while (TRUE) >> { >> /* >> ** Check to see if the child process has created the synchronizing >> logical >> ** name; If it has, then exit the loop. >> */ >> status = SYS$TRNLNM (&SyncLnmAttr, >> &SyncTblDesc, >> &SyncLnmDesc, >> 0, 0); >> if (status & 1) >> break; >> [cut] >> >> The sync logical is defined within apache$dcl.com and the above >> routine just runs in a tight loop until the definition takes place. >> >> At times this can take 30% of a horrendous number of logical name >> translations. >> >> John > > Is there some reason why this code can't have a built in wait? Waiting > for two or five seconds between checks would save one hell of a lot of > CPU that could probably be better used elsewhere. Richard, I suppose it could, and wondered about that myself, however, the offending code is part of the Apache web server. Sorry for not making that more clear. John ------------------------------ Date: Tue, 11 Mar 2008 14:40:02 -0600 From: John Nebel Subject: Re: Apache performance problem Message-ID: <47D6EE22.5070904@csdco.com> VAXman-@SendSpamHere.ORG wrote: > In article <47D6E7DA.1060107@comcast.net>, "Richard B. Gilbert" writes: >> John Nebel wrote: >>> >>> Alan Winston - SSRL Central Computing wrote: >>> >>>> In article <47D045E6.7050303@csdco.com>, John Nebel >>>> writes: >>>> >>>>> This is on OpenVMS 7.3-2 and Apache 1.3-1 >>>>> >>>>> Apache is performing an insane number of logical name translations, >>>>> this can exceed a rate of 200k/second and averages around 10k/second >>>>> in our environment. >>>>> >>>>> One thing Apache appears to be doing is looping on a translation of >>>>> apache$dcl_sync_xxxxxxxx >>>>> >>>>> 6-MAR 11:55:27.182282 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> 6-MAR 11:55:27.182266 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> 6-MAR 11:55:27.182250 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> 6-MAR 11:55:27.182235 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> >>>>> Ultimately this appears to have the ability to jam VMS enough to >>>>> crash it with a cluexit. >>>>> >>>>> Is their a workaround, or does Apache 2.1-1 have a more reasonable >>>>> behavior? >>>> >>>> >>>> There's a CRTL flag that'll make Apache cache those logical-name >>>> translation >>>> results. >>>> $ define/nolog decc$enable_getenv_cache enable >>>> >>>> in the APACHE$WWW LOGIN.COM >>>> >>>> -- Alan >>> >>> Alan, >>> >>> I don't think the decc$enable_getenv_cache logical will help in this >>> case - the offending code within Apache (proc.c) starts with: >>> >>> ** Loop here until the child creates the synchronizing logical name or >>> until >>> ** the child process has exited. >>> */ >>> while (TRUE) >>> { >>> /* >>> ** Check to see if the child process has created the synchronizing >>> logical >>> ** name; If it has, then exit the loop. >>> */ >>> status = SYS$TRNLNM (&SyncLnmAttr, >>> &SyncTblDesc, >>> &SyncLnmDesc, >>> 0, 0); >>> if (status & 1) >>> break; >>> [cut] >>> >>> The sync logical is defined within apache$dcl.com and the above routine >>> just runs in a tight loop until the definition takes place. >>> >>> At times this can take 30% of a horrendous number of logical name >>> translations. >>> >>> John >> Is there some reason why this code can't have a built in wait? Waiting >> for two or five seconds between checks would save one hell of a lot of >> CPU that could probably be better used elsewhere. > > This is FN ugly! Is there some reason _why_ this has been implemented in > this fashion? There are better ways to inform the parent process that a > subprocess has been created instead of looping on a semaphore implemented > with a logical name. > > It produces interesting results when a shadow merge is happening on the disk containing sysuaf - think locked up solid! Other times, I believe, setting up a sequence of events leading to cluexit. John ------------------------------ Date: Wed, 12 Mar 2008 03:45:33 GMT From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) Subject: Re: Apache performance problem Message-ID: <00A76738.EEA9033D@SSRL.SLAC.STANFORD.EDU> In article <47d6ea50$0$25052$607ed4bc@cv.net>, VAXman- @SendSpamHere.ORG writes: >In article <47D6E7DA.1060107@comcast.net>, "Richard B. Gilbert" writes: >>John Nebel wrote: >>> >>> >>> Alan Winston - SSRL Central Computing wrote: >>> >>>> In article <47D045E6.7050303@csdco.com>, John Nebel >>>> writes: >>>> >>>>> This is on OpenVMS 7.3-2 and Apache 1.3-1 >>>>> >>>>> Apache is performing an insane number of logical name translations, >>>>> this can exceed a rate of 200k/second and averages around 10k/second >>>>> in our environment. >>>>> >>>>> One thing Apache appears to be doing is looping on a translation of >>>>> apache$dcl_sync_xxxxxxxx >>>>> >>>>> 6-MAR 11:55:27.182282 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> 6-MAR 11:55:27.182266 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> 6-MAR 11:55:27.182250 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> 6-MAR 11:55:27.182235 03 4DF1DF06 APACHE_HTTPD 000FA418 >>>>> APACHE$DCL_SYNC_4DFAA98B >>>>> >>>>> Ultimately this appears to have the ability to jam VMS enough to >>>>> crash it with a cluexit. >>>>> >>>>> Is their a workaround, or does Apache 2.1-1 have a more reasonable >>>>> behavior? >>>> >>>> >>>> >>>> There's a CRTL flag that'll make Apache cache those logical-name >>>> translation >>>> results. >>>> $ define/nolog decc$enable_getenv_cache enable >>>> >>>> in the APACHE$WWW LOGIN.COM >>>> >>>> -- Alan >>> >>> >>> Alan, >>> >>> I don't think the decc$enable_getenv_cache logical will help in this >>> case - the offending code within Apache (proc.c) starts with: >>> >>> ** Loop here until the child creates the synchronizing logical name or >>> until >>> ** the child process has exited. >>> */ >>> while (TRUE) >>> { >>> /* >>> ** Check to see if the child process has created the synchronizing >>> logical >>> ** name; If it has, then exit the loop. >>> */ >>> status = SYS$TRNLNM (&SyncLnmAttr, >>> &SyncTblDesc, >>> &SyncLnmDesc, >>> 0, 0); >>> if (status & 1) >>> break; >>> [cut] >>> >>> The sync logical is defined within apache$dcl.com and the above routine >>> just runs in a tight loop until the definition takes place. >>> >>> At times this can take 30% of a horrendous number of logical name >>> translations. >>> >>> John >> >>Is there some reason why this code can't have a built in wait? Waiting >>for two or five seconds between checks would save one hell of a lot of >>CPU that could probably be better used elsewhere. It would certainly save CPU, but at the cost of delaying response to the client by 2-5 seconds when a new child needed to be created. > >This is FN ugly! Is there some reason _why_ this has been implemented in >this fashion? There are better ways to inform the parent process that a >subprocess has been created instead of looping on a semaphore implemented >with a logical name. Create a mailbox for the child, have the child write to it first thing, and hang a read attention AST on it, for example. I presume the idea is to not have to completely restructure the original Apache code, though. (But I don't pretend to speak for the CSWS developers.) -- Alan ------------------------------ Date: Wed, 12 Mar 2008 00:38:55 GMT From: David Turner Subject: Cheap Gigabit cards at Island Computers Message-ID: Refurbished 3x-DEGXA-TA with 1yr warranty in stock in Qty for only $299 each Buy Now!!!!!!!!!!!!!!!!!!!!!!!!!!! David Turner www.islandco.com dturner@islandco.com ------------------------------ Date: Wed, 12 Mar 2008 00:13:06 -0500 From: "Schroeder, AJ" Subject: Re: DHCP Server question/problem Message-ID: <47d76665$0$4941$4c368faf@roadrunner.com> JF Mezei wrote: > Peter Weaver wrote: > >> Finally I gave up on fixing the Alpha and decided to let my router >> do the DHCP serving, so I enabled the DHCP serving and rebooted the >> router. As the router was rebooting the two devices both received >> their DHCP information from the Alpha. So something in my router >> decided to block the DHCP response. > > DHCP is a LAN level protocol. It was not originally designed to be > routable. Extensions have later permitted routers to pass DHCP > requests and responses across a router. > > In my case, the Mac receives the DHCP response and gets an IP, But it > doesn't receive or process any additional information such as the > default route (gateway router) or the list of DNS servers available. Well, I am not a VMS expert (I didn't even know that they had a DHCP server!) Anyway, I do have quite a bit of DHCP experience. Whenever I ran into devices that didn't seem to accept certain options I narrowed it down to the device not understanding certain options in the DHCPOFFER stage of the process. Per RFC there are certain numerical options (3 for default router, 6 for DNS server,15 for domain name, etc.) My money would be that VMS isn't passing the default router and the DNS server in such a way as the Macs expect, so the Macs are just ignoring those options. Have you hooked the Mac "back to back" with your Alpha DHCP server with a hub and gotten a sniff trace? I assume that you have a broadband connection, can you plug one of your Macs into the DSL/Cable modem and/or your router to see if you get an 1) get an address and 2) verify that you have a default router and DNS settings? Just some things to try... ------------------------------ Date: Tue, 11 Mar 2008 11:13:14 -0700 (PDT) From: Neil Rieck Subject: Re: from TheDailyWTF Message-ID: <67489d4e-2b08-4aa6-af42-cc27d54c8b29@e23g2000prf.googlegroups.com> On Mar 5, 8:55=A0pm, Arne Vajh=F8j wrote: > My apologies if it has been posted before: > > http://thedailywtf.com/Articles/Jurassic-Programmers-.aspx > > Arne This article was great. Thanks. * * * The first time Nortel almost went under in the late 1980s, some pointy- haired boss decided that it was too hard to train new employees how to program in PROTEL (a variant of Pascal) which was the s/w running their DMS telephone switches. Nortel then embarked upon a massive rewritting of the DMS software from PROTEL to C. While they were doing this, they ran into many s/w problems and weren't able to ship anything new for about 18 months. (This was the time that their competition ate them for lunch). I guess I don't need to mention that their customers didn't care what language was used in the switch. They only wanted a product that worked. Neil Rieck Kitchener/Waterloo/Cambridge, Ontario, Canada. http://www3.sympatico.ca/n.rieck/ ------------------------------ Date: Wed, 12 Mar 2008 02:19:42 GMT From: Malcolm Dunnett Subject: Re: OpenVMS Oracle v7.3.3 to Oracle 10GR2 migration and application development Message-ID: <47D73DC0.10704@spammers.are.scum> Michael Austin wrote: > > Just curious, but, how do you propose to get from 7.3.3 to 10g? min > versions that can go directly to 10g are 8.1.7.4 or 9.2.0.4 Export/Import? re linking: Generally the only thing you should need to link to with 10g is libclntsh.so If you are using OCI calls be aware that there's a completely new OCI model starting with Oracle 8 (though the V7 calls are supported for compatibility) ------------------------------ Date: Tue, 11 Mar 2008 11:14:04 -0700 (PDT) From: Doug Phillips Subject: Re: Proof that macintosh is better than VMS Message-ID: <4322f95d-b047-405b-9d2d-0e25889df10a@s37g2000prg.googlegroups.com> On Mar 11, 2:16 pm, koeh...@eisner.nospam.encompasserve.org (Bob Koehler) wrote: > In article <960d254f-6ae7-4334-ab8e-e58e2b1ed...@8g2000hse.googlegroups.com>, Doug Phillips writes: > > > > > You are confusing quantum mechanics math with reality. If you mean > > that the mathematics of quantum mechanics is not concerned with > > resolving apparent randomness, then you are correct. You might want to > > look into the de Broglie-Bohm theory, more recently called Bohmian > > Mechanics. > > Quantum mechanics math vs. reality? You think reality differs? I believe that reality does consist of probabilities, but QM math is vague by design and accepts those probabilities as "good enough" because in most cases they are. If you believe that Schrodinger's Cat was really both 1/2 alive and 1/2 dead, then you are welcome to that belief. If you believe that the cat is either dead or alive and the experiment simply illustrates a measurement problem, then you're getting closer to what I believe. I suggest reading what John Bell has said about some of the problems with QM. If you want to advance further beyond the world of dead- scientists, then read some of the work being done by people who are still alive and with whom you can actually discuss this. You're not too far from Rutger's, are you (or am I thinking of someone else)? Look up Sheldon Goldstein or one of his contemporaries and discuss it with them if you feel qualified. You might find that Schrodinger's, Einstein's, Bell's, de Broglie's, Bohm's (and other "non-conformist") thoughts are being taken more seriously these days. You'll just have to look into it yourself. I'm not going to get into a battle of dead-scientists. ------------------------------ Date: 11 Mar 2008 18:19:52 GMT From: billg999@cs.uofs.edu (Bill Gunshannon) Subject: Re: Proof that macintosh is better than VMS Message-ID: <63o0q8F27i0npU1@mid.individual.net> In article , koehler@eisner.nospam.encompasserve.org (Bob Koehler) writes: > In article <960d254f-6ae7-4334-ab8e-e58e2b1ed88c@8g2000hse.googlegroups.com>, Doug Phillips writes: >> >> You are confusing quantum mechanics math with reality. If you mean >> that the mathematics of quantum mechanics is not concerned with >> resolving apparent randomness, then you are correct. You might want to >> look into the de Broglie-Bohm theory, more recently called Bohmian >> Mechanics. > > Quantum mechanics math vs. reality? You think reality differs? I'll bet a lot of people do. When science requires faith than religion in order to accept that which can neither be observed nor satisfactorily proven I think more and more people will see the difference. bill -- Bill Gunshannon | de-moc-ra-cy (di mok' ra see) n. Three wolves billg999@cs.scranton.edu | and a sheep voting on what's for dinner. University of Scranton | Scranton, Pennsylvania | #include ------------------------------ Date: 11 Mar 2008 17:20:56 -0600 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: Proof that macintosh is better than VMS Message-ID: In article <63o0q8F27i0npU1@mid.individual.net>, billg999@cs.uofs.edu (Bill Gunshannon) writes: > In article , > koehler@eisner.nospam.encompasserve.org (Bob Koehler) writes: >> In article <960d254f-6ae7-4334-ab8e-e58e2b1ed88c@8g2000hse.googlegroups.com>, Doug Phillips writes: >>> >>> You are confusing quantum mechanics math with reality. If you mean >>> that the mathematics of quantum mechanics is not concerned with >>> resolving apparent randomness, then you are correct. You might want to >>> look into the de Broglie-Bohm theory, more recently called Bohmian >>> Mechanics. >> >> Quantum mechanics math vs. reality? You think reality differs? > > I'll bet a lot of people do. When science requires faith than religion > in order to accept that which can neither be observed nor satisfactorily > proven I think more and more people will see the difference. > All of the predictions of quantum mechanics have been observed and verified many times. If they weren't the fellow who observed the failure would get the Nobel Prize. ------------------------------ Date: Tue, 11 Mar 2008 17:48:03 -0700 (PDT) From: AEF Subject: Re: Proof that macintosh is better than VMS Message-ID: On Mar 11, 1:19 pm, billg...@cs.uofs.edu (Bill Gunshannon) wrote: > In article , > koeh...@eisner.nospam.encompasserve.org (Bob Koehler) writes: > > > In article <960d254f-6ae7-4334-ab8e-e58e2b1ed...@8g2000hse.googlegroups.com>, Doug Phillips writes: > > >> You are confusing quantum mechanics math with reality. If you mean > >> that the mathematics of quantum mechanics is not concerned with > >> resolving apparent randomness, then you are correct. You might want to > >> look into the de Broglie-Bohm theory, more recently called Bohmian > >> Mechanics. > > > Quantum mechanics math vs. reality? You think reality differs? > > I'll bet a lot of people do. When science requires faith than religion > in order to accept that which can neither be observed nor satisfactorily > proven I think more and more people will see the difference. I assume you meant "When science requires *more* faith..." Scientists have faith in the scientific method which requires evidence. Religious people have what James Randi calls "blind faith"[1]. That makes all the difference in the world. [1] See http://www.randi.org/jr/072503.html (Mostly a good article, but I disagree with his opinion of the Wizard of Oz.) As far as using local hidden variables to restore determinism that only "appears" probabilistic, the experimental evidence ruling these out is more compelling than ever. Many, many experiments have been done and QM always, always wins. We're not talking about the possibility of experimental error clouding the results. The skeptics who complained that the early experiments could still allow local hidden variables because of events missed by detectors because said detectors were not 100% efficient. OK. But the efficiencies have been greatly improved and the room for determinism has been all but wiped out. Then there is the GHZ paradox which largely sidesteps the issue. There is simply no way to explain the results of GHZ experiments using local hidden variables. If you would learn about this, you would probably slowly begin to realize that there is no way out. It is getting to the point where insisting there must be determinism somehow being hidden behind the veil of probability is akin to denying the existence of atoms. Do you deny the existence of atoms? If so, why; and if not, why not? I used to be on the deterministic side. I even tried to concoct an explanation for polarization experiments to show how determinism could still prevail, but I quickly found my analysis to be flawed. The only faith science requires is faith in evidence, which is exactly the opposite of religious faith, which is faith that some people hundreds and/or thousands of years ago interacted with some god and wrote about it. That's hardly the same. AEF > > bill > > -- > Bill Gunshannon | de-moc-ra-cy (di mok' ra see) n. Three wolves > billg...@cs.scranton.edu | and a sheep voting on what's for dinner. > University of Scranton | > Scranton, Pennsylvania | #include ------------------------------ Date: Tue, 11 Mar 2008 17:57:33 -0700 (PDT) From: AEF Subject: Re: Proof that macintosh is better than VMS Message-ID: <969c8266-38a4-4a80-a9f4-711e65b0f4ff@f63g2000hsf.googlegroups.com> On Mar 11, 11:35 am, Doug Phillips wrote: > On Mar 11, 9:23 am, koeh...@eisner.nospam.encompasserve.org (Bob > > > > Koehler) wrote: > > In article <47d58fcf$0$1443$c3e8...@news.astraweb.com>, JF Mezei writes: > > > > However, prior to the draw, even if you knew the full configuration of > > > the machine, balls and aerodynamic properties of the chamber and the fan > > > blowing in it, you cannot predict the exact position of the rotating > > > chamber at time of start, the exact time difference between start of > > > rotation and the moment they drop the balls into the chamber, and the > > > exact moment when some human pushes a button to get a ball to come out. > > > So even if physics, aerodynamics and others sciences can explain the > > > movement of balls in the machine, no human has sufficient information to > > > have all the variables and thus, the outcome is random at the human level. > > > The random outcome of quantum mechanical behaviour is not a mere > > limitation of human capability. > > You are confusing quantum mechanics math with reality. If you mean > that the mathematics of quantum mechanics is not concerned with > resolving apparent randomness, then you are correct. You might want to > look into the de Broglie-Bohm theory, more recently called Bohmian > Mechanics. It seems to me that the apparent randomness isn't really addressed by this theory. In fact, I searched for "random" in the Wikipedia article about it and got zero hits. The theory attempts to show that particles have definite positions and velocities even during their wave-like interactions. But these are unobservable and still are subject to random movement. The theory, AFAICT, attempts to restore "realism". But I have become a positivist -- one who says that if it can't be observed, what does it matter. QM is astonishingly successful in predicting what we can measure. What happens in between -- what does it matter? I like to bring up the idea of parallel universes that never interact with ours so we can never observe any effects of them. If there are no testable predictions, one can't check it anyway, so why bother? (Reminds me of string theory!) AEF ------------------------------ Date: Tue, 11 Mar 2008 18:03:04 -0700 (PDT) From: AEF Subject: Re: Proof that macintosh is better than VMS Message-ID: <4537ae9f-46da-448d-b7cf-a5b77f6a9313@x41g2000hsb.googlegroups.com> On Mar 11, 1:14 pm, Doug Phillips wrote: > On Mar 11, 2:16 pm, koeh...@eisner.nospam.encompasserve.org (Bob > > Koehler) wrote: > > In article <960d254f-6ae7-4334-ab8e-e58e2b1ed...@8g2000hse.googlegroups.com>, Doug Phillips writes: > > > > You are confusing quantum mechanics math with reality. If you mean > > > that the mathematics of quantum mechanics is not concerned with > > > resolving apparent randomness, then you are correct. You might want to > > > look into the de Broglie-Bohm theory, more recently called Bohmian > > > Mechanics. > > > Quantum mechanics math vs. reality? You think reality differs? > > I believe that reality does consist of probabilities, but QM math is > vague by design and accepts those probabilities as "good enough" > because in most cases they are. QM math is not vague. The wave functions are well determined and so are the probabilities for the various outcomes therefrom. > > If you believe that Schrodinger's Cat was really both 1/2 alive and > 1/2 dead, then you are welcome to that belief. If you believe that the > cat is either dead or alive and the experiment simply illustrates a > measurement problem, then you're getting closer to what I believe. I, for one, do not believe the cat is half dead or half alive. The cat is a macroscopic object. But when a particle (be it a photon, electron, or what have you) passes through a microscopic double slit when the appartus is set up in a way that produces an interference pattern, it is as if it passes through both slits at the same time. Such superpositions of even atoms have been achieved within the last few years or so (maybe even a decade or more -- time passes so fast as you age!). > > I suggest reading what John Bell has said about some of the problems > with QM. If you want to advance further beyond the world of dead- > scientists, then read some of the work being done by people who are > still alive and with whom you can actually discuss this. OK. Can you suggest specific references? > > You're not too far from Rutger's, are you (or am I thinking of someone > else)? Look up Sheldon Goldstein or one of his contemporaries and > discuss it with them if you feel qualified. You mean pay him a visit? > > You might find that Schrodinger's, Einstein's, Bell's, de Broglie's, > Bohm's (and other "non-conformist") thoughts are being taken more > seriously these days. You'll just have to look into it yourself. I'm > not going to get into a battle of dead-scientists. Really? I don't see how in light of increasingly better experiments, in particular, the GHZ experiment. AEF ------------------------------ Date: Tue, 11 Mar 2008 18:16:39 -0700 (PDT) From: AEF Subject: Re: Proof that macintosh is better than VMS Message-ID: <2a2982eb-bc76-44c9-b17c-5070db7ef3fb@a1g2000hsb.googlegroups.com> On Mar 11, 6:20 pm, koeh...@eisner.nospam.encompasserve.org (Bob Koehler) wrote: > In article <63o0q8F27i0n...@mid.individual.net>, billg...@cs.uofs.edu (Bill Gunshannon) writes: > > In article , > > koeh...@eisner.nospam.encompasserve.org (Bob Koehler) writes: > >> In article <960d254f-6ae7-4334-ab8e-e58e2b1ed...@8g2000hse.googlegroups.com>, Doug Phillips writes: > > >>> You are confusing quantum mechanics math with reality. If you mean > >>> that the mathematics of quantum mechanics is not concerned with > >>> resolving apparent randomness, then you are correct. You might want to > >>> look into the de Broglie-Bohm theory, more recently called Bohmian > >>> Mechanics. > > >> Quantum mechanics math vs. reality? You think reality differs? > > > I'll bet a lot of people do. When science requires faith than religion > > in order to accept that which can neither be observed nor satisfactorily > > proven I think more and more people will see the difference. > > All of the predictions of quantum mechanics have been observed and > verified many times. If they weren't the fellow who observed the > failure would get the Nobel Prize. Exactly. Almost all of our modern technology wouldn't work otherwise. Transistors and lasers come to mind. Especially lasers. The idea of the laser is very, very QM. In fact, the idea was so strange that Einstein -- the one who first came up with the idea of spontaneous emission -- found it so strange that even he hesitated to tell anyone (or so I recall -- I could be wrong on this). Also, the photoelectric effect, black body radiation, the stability of atoms, Compton scattering, and many other phenomena are completely unexplainable without QM. In fact, classical physics (pre-quantum, or pre-1900), predicts completely different things. According to electromagnetic theory, electrons orbiting an atomic nucleus theory would radiate. This would cause the electrons to lose energy and thereby spiral into the nucleus in about 10**(-10) seconds (one-tenth of a billionth of a second). QM saves the day. Classical theory predicts completely different results for the other phenomena I mentioned. Additionally, there are paradoxes in statistical physics such as the Rayleigh-Jeans law for the energy density of electromagnetic radiation in thermal equilibrium is predicted to be infinite when summed up over all possible frequencies! And then there's the Gibbs paradox. QM comes to the rescue for both. Please read Chapter 6 of The Character of Physical Law (it's not a difficult read) and let me know if you can find any fault with Feynman's argument against hidden variable theories. If you don't want to splurge the $16 for the book, please watch part I of the 1979 Feynman lecture in Aukland as he shows that photons reflect probabilistically off of glass and water. AEF AEF ------------------------------ Date: Tue, 11 Mar 2008 18:41:54 -0700 (PDT) From: AEF Subject: Re: Proof that macintosh is better than VMS Message-ID: <2fdb5c8a-8452-47cb-9afe-5d2d0d28bd1e@n36g2000hse.googlegroups.com> On Mar 11, 1:19 pm, billg...@cs.uofs.edu (Bill Gunshannon) wrote: > In article , > koeh...@eisner.nospam.encompasserve.org (Bob Koehler) writes: > > > In article <960d254f-6ae7-4334-ab8e-e58e2b1ed...@8g2000hse.googlegroups.com>, Doug Phillips writes: > > >> You are confusing quantum mechanics math with reality. If you mean > >> that the mathematics of quantum mechanics is not concerned with > >> resolving apparent randomness, then you are correct. You might want to > >> look into the de Broglie-Bohm theory, more recently called Bohmian > >> Mechanics. > > > Quantum mechanics math vs. reality? You think reality differs? > > I'll bet a lot of people do. When science requires faith than religion > in order to accept that which can neither be observed nor satisfactorily > proven I think more and more people will see the difference. > > bill > > -- > Bill Gunshannon | de-moc-ra-cy (di mok' ra see) n. Three wolves > billg...@cs.scranton.edu | and a sheep voting on what's for dinner. > University of Scranton | > Scranton, Pennsylvania | #include Just one more word on this: Yes, it is true that no theory can be proved 100% correct. A theory can only be proved wrong. But in some cases, the evidence is so compelling that, for all practical purposes, it becomes accepted fact. Consider atomic theory. Do you doubt that ordinary matter is made of atoms? It took a lot of work to show that this was the case. The clincher was the discovery of Brownian motion and the explanation of it given by Einstein in 1905. After that, very few scientists doubted the reality of atoms (because there was no other reasonable way to explain the phenomenon). Other parts of science are less certain -- mostly the frontiers on the cutting edge of research. And it is good that people are skeptical here. As Feynman says in the book I've been referencing over and over: "... we are trying to prove ourselves wrong as quickly as possible, because only in that way can we find progress." For example, is CPT really a true, never-violated symmetry in nature? [Slight digression -- an explanation of CPT: It was once thought that parity (P) was conserved (not proved, but it was a quite reasonable conjecture at the time), but certain reactions involving the weak nuclear force violated it (kaon decay), sometimes even to the maximum possible extent (beta decay). Then it was postulated that CP (charge conjugation and parity together) was conserved, but the weak force violated even that in some other reactions (I think involving particles containing bottom quarks). So now time reversal (T, which was also thought to be conserved separately) was added and now it is thought that all three together (CPT) is conserved in all possible cases, and so far no one has seen any violations, but one day maybe someone will find one.] But would you doubt that there are eight planets orbiting a very hot sun? Do you doubt that we live on the third planet from the sun? I bet not. Some things like the speed-of-light barrier, and the existence of atoms, are all but beyond any doubt. QM seems to be in the same camp, along with its lack of determinism (note that the wave function *is* deterministic). The odds of a reversal in these cases is akin to winning the jackpot in the lottery. That is why people like Bob Koehler and me write what we write with such confidence on this topic. Please read The Character of Physical Law. It is an absolutely brilliant book, and even as a Ph.D. in physics I have learned much from it. AEF ------------------------------ Date: Tue, 11 Mar 2008 18:46:44 -0700 (PDT) From: AEF Subject: Re: Proof that macintosh is better than VMS Message-ID: <9e0b1458-ceaa-41ae-9e45-7cc0f8e5d034@d62g2000hsf.googlegroups.com> On Mar 11, 8:16 pm, AEF wrote: [...] > Please read Chapter 6 of The Character of Physical Law (it's not a > difficult read) and let me know if you can find any fault with > Feynman's argument against hidden variable theories. If you don't want > to splurge the $16 for the book, please watch part I of the 1979 > Feynman lecture in Aukland as he shows that photons reflect > probabilistically off of glass and water. You can find the video at www.feynman.com. > > AEF > > AEF ------------------------------ Date: Tue, 11 Mar 2008 14:08:37 -0500 From: norm.raphael@metso.com Subject: Time and TDF after DST advance Message-ID: This is a multipart message in MIME format. --=_alternative 006928D985257409_= Content-Type: text/plain; charset="US-ASCII" I do not run DTSS. I do run Multinet. The time changes via NTP. The multinet _timezone logical changes without any action from me. I submitted Sys$Manager:Utc$Time_Setup.com to correct the TDF with "/AFTER=03:15". At 3:12 EDT a show time gave 02:12, but at 3:16 EDT the batch had run and a show time gave 03:16. That looks like the system knew what time it was, but the show time command did not. Can someone explain, pls. --=_alternative 006928D985257409_= Content-Type: text/html; charset="US-ASCII"
I do not run DTSS.  I do run Multinet.  The time changes
via NTP.  The multinet _timezone logical changes without
any action from me.

I submitted Sys$Manager:Utc$Time_Setup.com to correct the
TDF with "/AFTER=03:15".  At 3:12 EDT a show time gave 02:12,
but at 3:16 EDT the batch had run and a show time gave 03:16.

That looks like the system knew what time it was, but the show
time command did not.

Can someone explain, pls.
--=_alternative 006928D985257409_=-- ------------------------------ Date: Tue, 11 Mar 2008 12:51:06 -0700 (PDT) From: tadamsmar Subject: Re: Time and TDF after DST advance Message-ID: On Mar 11, 3:08=A0pm, norm.raph...@metso.com wrote: > I do not run DTSS. =A0I do run Multinet. =A0The time changes > via NTP. =A0The multinet _timezone logical changes without > any action from me. > > I submitted Sys$Manager:Utc$Time_Setup.com to correct the > TDF with "/AFTER=3D03:15". =A0At 3:12 EDT a show time gave 02:12, > but at 3:16 EDT the batch had run and a show time gave 03:16. > > That looks like the system knew what time it was, but the show > time command did not. NTP knew GMT, but show time was not based on a correct TDF untill "AFTER 03:15" since that when TDF was set. > > Can someone explain, pls. At 3:12 EDT, you had a -5 hour TDF. At 3:16 EDT, you had a -4 TDF. NTP always just serves up GMT and VMS adds whatever TDF it has to get its system time that you see via SHOW TIME. I use to run with NTP. IIRC, the safest procedure is to shut down NTP, set the TDF, then restart NTP, but I don't know of a specifc gotcha if you leave NTP running. If you want it to work closer to perfect, then use /AFTER=3D02:00 ------------------------------ Date: Tue, 11 Mar 2008 16:01:47 -0400 From: "Richard B. Gilbert" Subject: Re: Time and TDF after DST advance Message-ID: <47D6E52B.2070809@comcast.net> norm.raphael@metso.com wrote: > > I do not run DTSS. I do run Multinet. The time changes > via NTP. The multinet _timezone logical changes without > any action from me. > NTP has nothing to do with it or, a least, it shouldn't have anything to do with it! As far as NTP is concerned, timezones are a figment of your imagination. > I submitted Sys$Manager:Utc$Time_Setup.com to correct the > TDF with "/AFTER=03:15". At 3:12 EDT a show time gave 02:12, > but at 3:16 EDT the batch had run and a show time gave 03:16. > > That looks like the system knew what time it was, but the show > time command did not. > > Can someone explain, pls. ------------------------------ Date: Tue, 11 Mar 2008 13:09:24 -0700 (PDT) From: tadamsmar Subject: Re: Time and TDF after DST advance Message-ID: On Mar 11, 3:51=A0pm, tadamsmar wrote: > On Mar 11, 3:08=A0pm, norm.raph...@metso.com wrote: > > > I do not run DTSS. =A0I do run Multinet. =A0The time changes > > via NTP. =A0The multinet _timezone logical changes without > > any action from me. > > > I submitted Sys$Manager:Utc$Time_Setup.com to correct the > > TDF with "/AFTER=3D03:15". =A0At 3:12 EDT a show time gave 02:12, > > but at 3:16 EDT the batch had run and a show time gave 03:16. > > > That looks like the system knew what time it was, but the show > > time command did not. > > NTP knew GMT, but show time was not based on a correct TDF untill > "AFTER 03:15" since that when TDF was set. > > > > > Can someone explain, pls. > > At 3:12 EDT, =A0you had a -5 hour TDF. At 3:16 EDT, you had a -4 TDF. > NTP always just serves up GMT and VMS adds whatever TDF it has to get > its system time that you see via SHOW TIME. Actually, that explanation does not work. if you specified / AFTER=3D3:15 the the TDF was set at 3:15 EST, because the system was running on EST before, I assume. Is the multinet timezone change completely separate from VMS? As Gilbert said, NTP just serves up GMT and knows nothing about time zones nor daylight saving. VMS starts with GMT and creates its system time based on its TDF. > > I use to run with NTP. =A0IIRC, the safest procedure is to shut down > NTP, > set the TDF, then restart NTP, but I don't know of a specifc gotcha if > you leave NTP running. =A0If you want it to work closer to perfect, then > use /AFTER=3D02:00 ------------------------------ Date: 11 Mar 2008 17:25:24 -0600 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: Time and TDF after DST advance Message-ID: In article <47D6E52B.2070809@comcast.net>, "Richard B. Gilbert" writes: > norm.raphael@metso.com wrote: >> >> I do not run DTSS. I do run Multinet. The time changes >> via NTP. The multinet _timezone logical changes without >> any action from me. >> > > NTP has nothing to do with it or, a least, it shouldn't have anything to > do with it! As far as NTP is concerned, timezones are a figment of your > imagination. That is true on UNIX, but not on VMS. Multinet NTP does alter the system time to deal with the fact that VMS keeps the clock in local time. ------------------------------ Date: Tue, 11 Mar 2008 16:44:50 -0500 From: norm.raphael@metso.com Subject: Re: Time and TDF after DST advance Message-ID: This is a multipart message in MIME format. --=_alternative 0077764485257409_= Content-Type: text/plain; charset="US-ASCII" koehler@eisner.nospam.encompasserve.org (Bob Koehler) wrote on 03/11/2008 07:25:24 PM: > In article <47D6E52B.2070809@comcast.net>, "Richard B. Gilbert" > writes: > > norm.raphael@metso.com wrote: > >> > >> I do not run DTSS. I do run Multinet. The time changes > >> via NTP. The multinet _timezone logical changes without > >> any action from me. > >> > > > > NTP has nothing to do with it or, a least, it shouldn't have anything to > > do with it! As far as NTP is concerned, timezones are a figment of your > > imagination. > > That is true on UNIX, but not on VMS. Multinet NTP does alter the > system time to deal with the fact that VMS keeps the clock in local > time. > That still leaves the fact that the Queue Manager seems to know it's EDT, but the Show time does not. Trying to submit this for fall-back is even dicier, since there are two 2:00 a.m.s. Setting the differential at 2:00 a.m. EDT to EST will show *what* during the next hour for show time? --=_alternative 0077764485257409_= Content-Type: text/html; charset="US-ASCII"
koehler@eisner.nospam.encompasserve.org (Bob Koehler) wrote on 03/11/2008 07:25:24 PM:

> In article <47D6E52B.2070809@comcast.net>, "Richard B. Gilbert"
> <rgilbert88@comcast.net> writes:
> > norm.raphael@metso.com wrote:
> >>
> >> I do not run DTSS.  I do run Multinet.  The time changes
> >> via NTP.  The multinet _timezone logical changes without
> >> any action from me.
> >>
> >
> > NTP has nothing to do with it or, a least, it shouldn't have anything to
> > do with it!  As far as NTP is concerned, timezones are a figment of your
> > imagination.
>
>    That is true on UNIX, but not on VMS.  Multinet NTP does alter the
>    system time to deal with the fact that VMS keeps the clock in local
>    time.
>
That still leaves the fact that the Queue Manager seems to know it's EDT, but

the Show time does not.  

Trying to submit this for fall-back is even dicier, since there are
two 2:00 a.m.s.  Setting the differential at 2:00 a.m. EDT
to EST will show *what* during the next hour for show time?

--=_alternative 0077764485257409_=-- ------------------------------ Date: Tue, 11 Mar 2008 11:00:59 -0700 (PDT) From: Volker Halle Subject: Re: Time changing after reboot Message-ID: <21a34568-58d9-4e5a-8d00-22ad85865686@s19g2000prg.googlegroups.com> And $ DUMP SYS$SYSTEM:SYS$TIMEZONE.DAT before shutdown. Volker. ------------------------------ Date: Tue, 11 Mar 2008 14:11:06 -0400 From: "Jeff Goodwin" Subject: Re: Time changing after reboot Message-ID: <47d6cb3c$0$17370$4c368faf@roadrunner.com> "tadamsmar" wrote in message news:056fcdca-4c33-4a6b-b84f-5a53d09669b4@q78g2000hsh.googlegroups.com... >I have some systems set up for automatic time change. Everything > appeared to work fine on March 9, but... > > I just had to reboot one of them and it jumped forward one hour! > Then I rebooted another one and the same thing happened on that > system. > Check the last line of this file and verify you have the correct rule set: SYS$STARTUP:TDF$UTC_STARTUP.COM -Jeff ------------------------------ Date: Tue, 11 Mar 2008 11:18:48 -0700 (PDT) From: Volker Halle Subject: Re: Time changing after reboot Message-ID: <9fbeac11-87d3-4f04-bfa3-b8852164ab34@s13g2000prd.googlegroups.com> Look at the SYS$SYSTEM:SYS$TIMEZONE.DAT files on your systems. Dump the files and look at the quadword at offset 0x10: Virtual block number 1 (00000001), 512 (0200) bytes 00000000 00000000 00000000 00000000 ................ 000000 000C0A14 0001E240 00000008 61C46800 .h=C4a....@=E2...... 000010 ^^^^^^^^^^^^^^^^^ This is the TDF value quadword loaded into EXE$GQ_TDF during boot (by SYSINIT). SDA> exa exe$gq_tdf EXE$GQ_TDF: 00000008.61C46800 ".h=C4a...." SDA> exa/time exe$gq_tdf 17-NOV-1858 01:00:00.00 ! I'm running in CET-1 right now When I rebooted a V7.3-2 system at noon today, I also had the time jump by +06:00 - note we are at CET and the switch to DST is to happen on the last sunday in march ! And this was the message printed on the console during boot: %TDF-I-TDFSET, Summer time or standard time changeover - new SYS $TIMEZONE_DIFFERENTIAL=3D3600/old=3D-18000. When I now looked at the SYS$TIMEZONE.DAT files, I found one, which had been created during boot and contained at TDF of FFFFFFD6 1729F800 which amounts to 5 hours. I need to collect more information to be able to make any sense out of this... Volker. ------------------------------ Date: Tue, 11 Mar 2008 11:35:52 -0700 (PDT) From: tadamsmar Subject: Re: Time changing after reboot Message-ID: On Mar 11, 2:11=A0pm, "Jeff Goodwin" wrote: > "tadamsmar" wrote in message > > news:056fcdca-4c33-4a6b-b84f-5a53d09669b4@q78g2000hsh.googlegroups.com... > > >I have some systems set up for automatic time change. =A0 Everything > > appeared to work fine on March 9, but... > > > I just had to reboot one of them and it jumped forward one hour! > > Then I rebooted another one and the same thing happened on that > > system. > > Check the last line of this file and verify you have the correct rule set:= > > SYS$STARTUP:TDF$UTC_STARTUP.COM > > -Jeff Thanks! That fixed it. I don't see anything in the TZ-V0200 release notes about fixing that. How was one to know to edit that? I guess you only see this problem if you reboot during the window where the old and new rules disagree? If so, that is a nice little pitfall. ------------------------------ Date: Tue, 11 Mar 2008 11:49:07 -0700 (PDT) From: ewilts@ewilts.org Subject: Re: Time changing after reboot Message-ID: <3ff2fee5-ca4b-4188-a061-8a91dbbac39e@i29g2000prf.googlegroups.com> On Mar 11, 1:35 pm, tadamsmar wrote: > On Mar 11, 2:11 pm, "Jeff Goodwin" wrote: > > > "tadamsmar" wrote in message > > >news:056fcdca-4c33-4a6b-b84f-5a53d09669b4@q78g2000hsh.googlegroups.com... > > > >I have some systems set up for automatic time change. Everything > > > appeared to work fine on March 9, but... > > > > I just had to reboot one of them and it jumped forward one hour! > > > Then I rebooted another one and the same thing happened on that > > > system. > > > Check the last line of this file and verify you have the correct rule set: > > > SYS$STARTUP:TDF$UTC_STARTUP.COM This is the script I run every spring change and so far it hasn't failed me: $ RUN SYS$SYSTEM:SYSMAN SET ENVIRONMENT/CLUSTER DO @SYS$MANAGER:UTC$TIME_SETUP "" tdf -300 60 DO @SYS$MANAGER:UTC$TIME_SETUP "" rule DO DEF/SYS/EXEC PMDF_TIMEZONE CDT EXIT $exit .../Ed ------------------------------ Date: Tue, 11 Mar 2008 11:49:50 -0700 (PDT) From: Volker Halle Subject: Re: Time changing after reboot Message-ID: <5129ba40-96e7-46a7-a0c2-58c0a0f34b89@h11g2000prf.googlegroups.com> Running @UTC$TIME_SETUP after installing a TZ or TDF patch is always a good suggestion. Volker. ------------------------------ Date: Tue, 11 Mar 2008 12:00:19 -0700 (PDT) From: tadamsmar Subject: Re: Time changing after reboot Message-ID: <008e1db0-332a-43bf-86ea-35eba4b0f33f@e25g2000prg.googlegroups.com> On Mar 11, 2:49=A0pm, Volker Halle wrote: > Running @UTC$TIME_SETUP after installing a TZ or TDF patch is always a > good suggestion. > > Volker. Might be a good suggestion but: 1. The TZ patch release notes do not suggest it, and 2. There must be more to it than that, since I ran UTC$TIME_SETUP about a gazillion times today to set time time back one hour, and it did not fix TDF$UTC_STARTUP.COM. Turns out that one of my systems did have it fixed. I guess I must have run UTC$TIME_SETUP with the mystery arguments that you know and I don't on that system. ------------------------------ Date: Tue, 11 Mar 2008 12:02:47 -0700 (PDT) From: tadamsmar Subject: Re: Time changing after reboot Message-ID: <1384052b-34e5-49fb-a737-63376569331a@s13g2000prd.googlegroups.com> On Mar 11, 2:49=A0pm, ewi...@ewilts.org wrote: > On Mar 11, 1:35 pm, tadamsmar wrote: > > > > > > > On Mar 11, 2:11 pm, "Jeff Goodwin" wrote: > > > > "tadamsmar" wrote in message > > > >news:056fcdca-4c33-4a6b-b84f-5a53d09669b4@q78g2000hsh.googlegroups.com.= .. > > > > >I have some systems set up for automatic time change. =A0 Everything > > > > appeared to work fine on March 9, but... > > > > > I just had to reboot one of them and it jumped forward one hour! > > > > Then I rebooted another one and the same thing happened on that > > > > system. > > > > Check the last line of this file and verify you have the correct rule = set: > > > > SYS$STARTUP:TDF$UTC_STARTUP.COM > > This is the script I run every spring change and so far it hasn't > failed me: > > $ RUN SYS$SYSTEM:SYSMAN > =A0 =A0 =A0 =A0 SET ENVIRONMENT/CLUSTER > =A0 =A0 =A0 =A0 DO @SYS$MANAGER:UTC$TIME_SETUP "" tdf -300 60 > =A0 =A0 =A0 =A0 DO @SYS$MANAGER:UTC$TIME_SETUP "" rule > =A0 =A0 =A0 =A0 DO DEF/SYS/EXEC PMDF_TIMEZONE CDT > =A0 =A0 =A0 =A0 EXIT > $exit > > =A0 =A0.../Ed- Hide quoted text - > > - Show quoted text - Are you on 7.3.2? As of that rev, it can be automated by setting AUTO_DLIGHT_SAV =3D 1. ------------------------------ Date: Tue, 11 Mar 2008 13:18:58 -0700 (PDT) From: Jim Subject: Re: Time changing after reboot Message-ID: On Mar 11, 2:35=A0pm, tadamsmar wrote: > On Mar 11, 2:11=A0pm, "Jeff Goodwin" wrote: > > > "tadamsmar" wrote in message > > >news:056fcdca-4c33-4a6b-b84f-5a53d09669b4@q78g2000hsh.googlegroups.com...= > > > >I have some systems set up for automatic time change. =A0 Everything > > > appeared to work fine on March 9, but... > > > > I just had to reboot one of them and it jumped forward one hour! > > > Then I rebooted another one and the same thing happened on that > > > system. > > > Check the last line of this file and verify you have the correct rule se= t: > > > SYS$STARTUP:TDF$UTC_STARTUP.COM > > > -Jeff > > Thanks! > > That fixed it. > > I don't see anything in the TZ-V0200 release notes about fixing that. > > How was one to know to edit that? > You shouldn't have to - that file is updated when you use SYS $MANAGER:NET$CONFIGURE.COM to configure your time differential factor. You probably want to execute this now as it is possible that other things (SYS$SYSTEM:SYS$TIMEZONE_SRC.DAT perhaps) are also incorrect. After execution of NET$CONFIGURE to change the TDF you'll also want to execute the updated SYS$STARTUP:TDF$UTC_STARTUP.COM. ------------------------------ End of INFO-VAX 2008.143 ************************