1 INFO-VAX	Thu, 10 Jul 2003	Volume 2003 : Issue 378       Contents: autogen question Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: Clustering Problem Re: CSWS v1.3 abort  Re: CSWS v1.3 abort  Re: CSWS v1.3 abort  Re: DEClaser 5100 (LN09) help  Re: DEClaser 5100 (LN09) help  Re: Good code profilers on VMS? + Re: HP World: Why Alpha's Omega Makes Sense  Re: KVM switches! New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME % Re: New lexical function F$DELTA_TIME  Old Mailbox Problem  Re: OpenVMS I64, a proposal / Re: OpenVMS Technical Seminar Highlights (some) / Re: OpenVMS Technical Seminar Highlights (some) / Re: OpenVMS Technical Seminar Highlights (some) / Re: OpenVMS Technical Seminar Highlights (some) / Re: OpenVMS Technical Seminar Highlights (some) 4 Re: Opinions please: EVA and database journal policy Re: PDP-11 OS Release Dates  Pearl Users  Re: Pearl Users  PPPOE? When? How?  Re: PPPOE? When? How? & Re: problem with MONITOR CLUSTER (VPM) Re: Restoring an backup  Re: Rethinking  V.M.S  Re: Running VMS off CD System symbols Re: System symbols Re: System symbols; typos in SYS$COMMON:[SYSMGR]AGEN$NEW_NODE_DEFAULTS.TEMPLATE 5 Re: vms security model - does it still exist on IA64? & Re: XFC, IO Size and Backup - Take Two  F ----------------------------------------------------------------------  + Date: Thu, 10 Jul 2003 10:35:28 +0000 (UTC) P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply) Subject: autogen question $ Message-ID: <bejfhf$srp$2@online.de>  I On one system, I have some secondary page and swap files located off the  * system disk.  When running AUTOGEN, I get:  . RMS-E-FEX, file already exists, not superseded> %SET-E-READERR, error reading DISK$LABEL:[DIR]PAGEFILE_1.SYS;1* -SYSTEM-W-ACCONFLICT, file access conflict/ %RMS-E-FEX, file already exists, not superseded > %SET-E-READERR, error reading DISK$LABEL:[DIR]SWAPFILE_1.SYS;1* -SYSTEM-W-ACCONFLICT, file access conflict  + Is this expected?  Should I worry about it?    ------------------------------  % Date: Thu, 10 Jul 2003 12:59:42 +0100 " From: "Gary" <gmcd@totalise.co.uk> Subject: Clustering Problem / Message-ID: <bejkdk$5c0$1@newsg2.svr.pol.co.uk>    Hi,   K I am attempting to Cluster 2 VAXstation 4000's, one is a model 90 the other  a model 60.   L I have installed VMS 7.3 on each machine separately followed by the HobbyistH Licences for VAX/VMS and the Layered Product Set. I then installed TCPIPF [from the Hobbyist Disk] on each machine enabling me to connect to the' machines from my PC using Reflection X.   J I then executed CLUSTER_CONFIG.COM on each machine, to make the Model 60 aD LAN Cluster Server and the model 90 Cluster member. On rebooting the2 machines in any sequence I get the error message -  : %SYSINT-E, error mounting system device, status = 007280B4  4 on OPA0 during cluster formation of the second node.    G I looked error message up in the  OpenVMS Cluster Config Document [June K 2002]  and section  A.7.4.2.3 Mount Failures - would seem to suggest I need 3 to replace my drives with those that support TCQ???   K Is there a way round this problem? I've not Clustered VMS systems before so 3 any constructive advice would be extremely welcome.    Thanks in advance,   Gary.   A PS I've attached some additional information which may be of some I assistance - the boot sequence in this instance was the Model 90 then the 	 Model 60.    $  $ sh sys  E OpenVMS V7.3 on node GARYV1 10-JUL-2003 12:20:57.72 Uptime 0 00:05:25   2 Pid Process Name State Pri I/O CPU Page flts Pages  + 20200101 SWAPPER HIB 16 0 0 00:00:00.13 0 0 7 20200106 CLUSTER_SERVER HIB 12 11 0 00:00:00.05 206 315 1 20200107 CONFIGURE HIB 11 6 0 00:00:00.02 114 176 / 20200108 LANACP HIB 13 40 0 00:00:00.10 370 712 - 2020010A IPCACP HIB 10 7 0 00:00:00.02 99 178 . 2020010B ERRFMT HIB 8 21 0 00:00:00.07 146 2283 2020010C CACHE_SERVER HIB 16 6 0 00:00:00.01 84 135 - 2020010D OPCOM HIB 7 41 0 00:00:00.06 165 225 4 2020010E AUDIT_SERVER HIB 9 47 0 00:00:00.14 521 7784 2020010F JOB_CONTROL HIB 10 25 0 00:00:00.05 224 3766 20200110 QUEUE_MANAGER HIB 9 47 0 00:00:00.21 662 10319 20200111 SECURITY_SERVER HIB 10 22 0 00:00:00.18 712 1315 5 20200112 DNS$ADVER LEF 5 1034 0 00:00:00.69 2989 2242 4 20200113 LES$ACP_V30 HIB 8 104 0 00:00:00.18 208 286/ 20200114 NET$ACP HIB 4 40 0 00:00:00.17 671 778 + 20200115 REMACP HIB 8 8 0 00:00:00.03 87 57 1 20200116 NET$EVD HIB 4 15 0 00:00:00.20 1472 1116 3 20200117 DTSS$CLERK LEF 12 40 0 00:00:00.16 559 864 1 20200118 SMISERVER HIB 9 36 0 00:00:00.16 365 556 2 20200119 TP_SERVER HIB 10 20 0 00:00:00.05 205 3185 2020011A TCPIP$INETACP HIB 8 64 0 00:00:00.21 553 618 7 2020011B TCPIP$FTP_1 LEF 10 99 0 00:00:00.50 1967 694 N 0 2020011C SYSTEM CUR 4 138 0 00:00:00.39 1079 352
 $ sh dev d) Device Device Error Volume Free Trans Mnt ( Name Status Count Label Blocks Count Cnt DNFS0: Online 0 1 GARYV1$DKA100: Mounted 0 OVMSVAXSYS 1399815 222 1  GAZVX2$DKA0: Online 0  GAZVX2$DKA300: Online 0  GAZVX2$DKA400: Online 0  $  $  $ sh dev d/fulF Disk DNFS0:, device type Foreign disk type 7, is online, file-oriented device, 5 shareable, mailbox device, device is a template only. $ Error count 0 Operations completed 0# Owner process "" Owner UIC [SYSTEM] 6 Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G:R,W) Reference count 0 Default buffer size 512 ( Total blocks 8388608 Sectors per track 0' Total cylinders 0 Tracks per cylinder 0 I Disk GARYV1$DKA100:, device type RZ26B, is online, mounted, file-oriented F device, shareable, served to cluster via MSCP Server, error logging is enabled.' Error count 0 Operations completed 6527 # Owner process "" Owner UIC [SYSTEM] 6 Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G:R,W+ Reference count 143 Default buffer size 512 ) Total blocks 2050860 Sectors per track 57 + Total cylinders 2570 Tracks per cylinder 14 2 Volume label "OVMSVAXSYS" Relative volume number 0$ Cluster size 9 Transaction count 2220 Free blocks 1399815 Maximum files allowed 102543 Extend quantity 5 Mount count 1 8 Mount status System Cache name "_GARYV1$DKA100:XQPCACHE": Extent cache size 64 Maximum blocks in extent cache 139981= File ID cache size 64 Blocks currently in extent cache 139941 3 Quota cache size 0 Maximum buffers in FCP cache 818 ; Volume owner UIC [1,1] Vol Prot S:RWCD,O:RWCD,G:RWCD,W:RWCD I Volume Status: ODS-2, subject to mount verification, protected subsystems @ enabled, file high-water marking, write-through caching enabled.F Disk GAZVX2$DKA0:, device type RZ26L, is online, file-oriented device,: shareable, available to cluster, error logging is enabled.$ Error count 0 Operations completed 0# Owner process "" Owner UIC [SYSTEM] 6 Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G:R,W) Reference count 0 Default buffer size 512 ; Host name "GAZVX2" Host type, avail VAXstation 4000-60, yes G Disk GAZVX2$DKA300:, device type RZ25, is online, file-oriented device, : shareable, available to cluster, error logging is enabled.$ Error count 0 Operations completed 0# Owner process "" Owner UIC [SYSTEM] 6 Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G:R,W) Reference count 0 Default buffer size 512 ; Host name "GAZVX2" Host type, avail VAXstation 4000-60, yes K Disk GAZVX2$DKA400:, device type TOSHIBA CD-ROM XM-5401TA, is online, file- K oriented device, shareable, available to cluster, error logging is enabled. $ Error count 0 Operations completed 0# Owner process "" Owner UIC [SYSTEM] 6 Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G:R,W) Reference count 0 Default buffer size 512 ; Host name "GAZVX2" Host type, avail VAXstation 4000-60, yes    ------------------------------  % Date: Thu, 10 Jul 2003 09:18:39 -0400 * From: "Stanley F. Quayle" <stan@stanq.com> Subject: Re: Clustering Problem . Message-ID: <3F0D2F6F.13104.1719CE6@localhost>  $ On 10 Jul 2003 at 12:59, Gary wrote:E > On rebooting the machines in any sequence I get the error message -  > < > %SYSINT-E, error mounting system device, status = 007280B4  3 You can do an $EXIT %X007280B4 to get this message:   B   %MOUNT-F-VOLALRMNT, another volume of same label already mounted  E When you installed VMS 7.3 on both systems, you gave each volume the  F same label.  Only one volume with a given label can be mounted in the  cluster at a time...  
 --Stan Quayle  Quayle Consulting Inc.  
 ----------C Stanley F. Quayle, P.E. N8SQ  +1 614-868-1363  Fax: +1 614 868-1671 1 8572 North Spring Ct. NW, Pickerington, OH  43147 = Preferred address:  stan@stanq.com       http://www.stanq.com    ------------------------------  % Date: Thu, 10 Jul 2003 09:29:33 -0400 + From: Ken Robinson <kenrbnsn1@patmedia.net>  Subject: Re: Clustering Problem A Message-ID: <5.2.1.1.2.20030710092535.0567dcf8@mail.patmedia.net>   5 At 09:18 AM 7/10/2003 -0400, Stanley F. Quayle wrote:   % >On 10 Jul 2003 at 12:59, Gary wrote: G > > On rebooting the machines in any sequence I get the error message -  > > > > > %SYSINT-E, error mounting system device, status = 007280B4 > 4 >You can do an $EXIT %X007280B4 to get this message: > D >   %MOUNT-F-VOLALRMNT, another volume of same label already mounted > E >When you installed VMS 7.3 on both systems, you gave each volume the F >same label.  Only one volume with a given label can be mounted in the >cluster at a time...   C I thought that this error message would have been fixed already. I  C encountered in 1984 when personal from the local DIGITAL office in  H Piscataway, NJ were trying to boot our first 3 node cluster. (3 VAX 785 % machines with separate system disks).   K I've told this story at least two times at different Magic sessions at the   old DECUS symposia.   
 Ken Robinson     ------------------------------  % Date: Thu, 10 Jul 2003 16:11:13 +0200 , From: "Reinhard Eigner" <antispam@garnix.de> Subject: Re: Clustering Problem / Message-ID: <bejs5v$cp7$07$1@news.t-online.com>   3 > GARYV1$DKA100: Mounted 0 OVMSVAXSYS 1399815 222 1  > GAZVX2$DKA0: Online 0  > GAZVX2$DKA300: Online 0  > GAZVX2$DKA400: Online 0   < Didn't you have to change the allocation class of the disks?7 On our ES40 3-node cluster all the disks beginning with " $1$DKAxxx instead of the node name   bye  Reinhard   ------------------------------  % Date: Thu, 10 Jul 2003 15:21:36 +0100 " From: "Gary" <gmcd@totalise.co.uk> Subject: Re: Clustering Problem / Message-ID: <bejsoj$sv4$1@newsg4.svr.pol.co.uk>   > Stan, I hadn't noticed that. I changed the volume label and it( works!! I now have a 2 node VAX cluster.   Regards,   Gary.  $ sh sys/cluster  E OpenVMS V7.3 on node GARYV1 10-JUL-2003 15:17:06.92 Uptime 0 00:15:11   2 Pid Process Name State Pri I/O CPU Page flts Pages  + 20200101 SWAPPER HIB 16 0 0 00:00:00.20 0 0   7 20200106 CLUSTER_SERVER HIB 12 28 0 00:00:00.08 252 642   1 20200107 CONFIGURE HIB 8 18 0 00:00:00.03 169 237   / 20200108 LANACP HIB 13 41 0 00:00:00.12 383 567   - 2020010A IPCACP HIB 10 7 0 00:00:00.01 99 178   . 2020010B ERRFMT HIB 8 22 0 00:00:00.05 130 212  3 2020010C CACHE_SERVER HIB 16 6 0 00:00:00.02 84 135   - 2020010D OPCOM HIB 6 50 0 00:00:00.04 242 127   4 2020010E AUDIT_SERVER HIB 9 57 0 00:00:00.16 512 904  3 2020010F JOB_CONTROL HIB 8 31 0 00:00:00.08 224 376   7 20200110 QUEUE_MANAGER HIB 10 61 0 00:00:00.12 659 1028   9 20200111 SECURITY_SERVER HIB 10 22 0 00:00:00.17 758 1231   5 20200112 DNS$ADVER LEF 5 1036 0 00:00:00.71 2758 2240   4 20200113 LES$ACP_V30 HIB 8 104 0 00:00:00.19 207 286  / 20200114 NET$ACP HIB 4 40 0 00:00:00.17 684 843   + 20200115 REMACP HIB 8 8 0 00:00:00.02 87 57   1 20200116 NET$EVD HIB 6 15 0 00:00:00.26 1408 1156   3 20200117 DTSS$CLERK LEF 12 47 0 00:00:00.15 595 666   1 20200118 SMISERVER HIB 9 36 0 00:00:00.13 365 588   1 20200119 TP_SERVER HIB 9 59 0 00:00:00.06 205 318   5 2020011A TCPIP$INETACP HIB 8 71 0 00:00:00.23 558 625   8 2020011B TCPIP$FTP_1 LEF 10 100 0 00:00:00.46 1933 662 N  . 2020011D SYSTEM CUR 4 87 0 00:00:00.24 787 428  E OpenVMS V7.3 on node GAZVX2 10-JUL-2003 15:17:07.00 Uptime 0 00:13:02   2 Pid Process Name State Pri I/O CPU Page flts Pages  + 20400101 SWAPPER HIB 16 0 0 00:00:00.17 0 0   7 20400106 CLUSTER_SERVER CUR 13 11 0 00:00:00.13 208 314   2 20400107 CONFIGURE HIB 10 11 0 00:00:00.05 132 202  / 20400108 LANACP HIB 13 47 0 00:00:00.30 422 895   - 2040010A IPCACP HIB 10 7 0 00:00:00.03 99 178   . 2040010B ERRFMT HIB 7 36 0 00:00:00.13 146 228  3 2040010C CACHE_SERVER HIB 16 6 0 00:00:00.04 84 135   - 2040010D OPCOM HIB 7 42 0 00:00:00.10 236 119   5 2040010E AUDIT_SERVER HIB 10 48 0 00:00:00.36 534 823   4 2040010F JOB_CONTROL HIB 10 34 0 00:00:00.17 243 395  9 20400110 SECURITY_SERVER HIB 10 22 0 00:00:00.21 768 1331   5 20400111 DNS$ADVER LEF 5 1034 0 00:00:01.33 2733 2317   7 20400112 QUEUE_MANAGER HIB 10 53 0 00:00:00.31 729 1096   4 20400113 LES$ACP_V30 HIB 8 104 0 00:00:00.38 208 286  / 20400114 NET$ACP HIB 4 43 0 00:00:00.41 565 903   + 20400115 REMACP HIB 8 8 0 00:00:00.03 87 57   1 20400116 NET$EVD HIB 5 15 0 00:00:00.54 1367 1376   3 20400117 DTSS$CLERK LEF 12 47 0 00:00:00.33 566 830   1 20400118 SMISERVER HIB 9 36 0 00:00:00.27 328 611   1 20400119 TP_SERVER HIB 9 52 0 00:00:00.53 205 318   6 2040011A TCPIP$INETACP HIB 10 65 0 00:00:00.41 554 618  7 2040011B TCPIP$FTP_1 LEF 10 99 0 00:00:00.99 1915 773 N    $   5 "Stanley F. Quayle" <stan@stanq.com> wrote in message ( news:3F0D2F6F.13104.1719CE6@localhost...& > On 10 Jul 2003 at 12:59, Gary wrote:G > > On rebooting the machines in any sequence I get the error message -  > > > > > %SYSINT-E, error mounting system device, status = 007280B4 > 5 > You can do an $EXIT %X007280B4 to get this message:  > D >   %MOUNT-F-VOLALRMNT, another volume of same label already mounted > F > When you installed VMS 7.3 on both systems, you gave each volume theG > same label.  Only one volume with a given label can be mounted in the  > cluster at a time... >  > --Stan Quayle  > Quayle Consulting Inc. >  > ----------E > Stanley F. Quayle, P.E. N8SQ  +1 614-868-1363  Fax: +1 614 868-1671 3 > 8572 North Spring Ct. NW, Pickerington, OH  43147 ? > Preferred address:  stan@stanq.com       http://www.stanq.com  >  >    ------------------------------  + Date: Thu, 10 Jul 2003 15:04:50 +0000 (UTC) - From: lewis@spyder.mitre.org (Keith A. Lewis)  Subject: Re: Clustering Problem . Message-ID: <bejvai$edr$2@newslocal.mitre.org>   Ken Robinson <kenrbnsn1@patmedia.net> writes in article <5.2.1.1.2.20030710092535.0567dcf8@mail.patmedia.net> dated Thu, 10 Jul 2003 09:29:33 -0400:6 >At 09:18 AM 7/10/2003 -0400, Stanley F. Quayle wrote: > & >>On 10 Jul 2003 at 12:59, Gary wrote:H >> > On rebooting the machines in any sequence I get the error message - >> >? >> > %SYSINT-E, error mounting system device, status = 007280B4  >>5 >>You can do an $EXIT %X007280B4 to get this message:  >>E >>   %MOUNT-F-VOLALRMNT, another volume of same label already mounted  >>F >>When you installed VMS 7.3 on both systems, you gave each volume theG >>same label.  Only one volume with a given label can be mounted in the  >>cluster at a time... > D >I thought that this error message would have been fixed already. I D >encountered in 1984 when personal from the local DIGITAL office in I >Piscataway, NJ were trying to boot our first 3 node cluster. (3 VAX 785  & >machines with separate system disks).  L Fixed??  It aint broke!  The only possible improvement I can see would be toK add the text corresponding to the status code.  The diagnostic is correct.  I You can't cluster systems booted from multiple system disks with the same > volume name because names of mounted volumes must be unique.    + --Keith Lewis              klewis$mitre.org > The above may not (yet) represent the opinions of my employer.   ------------------------------  % Date: Thu, 10 Jul 2003 11:08:15 -0400 + From: Ken Robinson <kenrbnsn1@patmedia.net>  Subject: Re: Clustering Problem A Message-ID: <5.2.1.1.2.20030710110511.052bfaa8@mail.patmedia.net>   2 At 03:04 PM 7/10/2003 +0000, Keith A. Lewis wrote: > > E > >I thought that this error message would have been fixed already. I E > >encountered in 1984 when personal from the local DIGITAL office in J > >Piscataway, NJ were trying to boot our first 3 node cluster. (3 VAX 785( > >machines with separate system disks). > M >Fixed??  It aint broke!  The only possible improvement I can see would be to K >add the text corresponding to the status code.  The diagnostic is correct. J >You can't cluster systems booted from multiple system disks with the same= >volume name because names of mounted volumes must be unique.   K That's what I meant by "fixed".  Once you know the meaning of the message,  K you say "of course", but until then you can go crazy. Back in 1984 it took  J DIGITAL almost two weeks to figure it out ... of course clusters were new - then and there wasn't as much expertise then.    Ken    ------------------------------  + Date: Thu, 10 Jul 2003 15:48:51 +0000 (UTC) P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply) Subject: Re: Clustering Problem $ Message-ID: <bek1t3$c8m$1@online.de>  A In article <bejs5v$cp7$07$1@news.t-online.com>, "Reinhard Eigner"  <antispam@garnix.de> writes:    5 > > GARYV1$DKA100: Mounted 0 OVMSVAXSYS 1399815 222 1  > > GAZVX2$DKA0: Online 0  > > GAZVX2$DKA300: Online 0  > > GAZVX2$DKA400: Online 0  > > > Didn't you have to change the allocation class of the disks?9 > On our ES40 3-node cluster all the disks beginning with $ > $1$DKAxxx instead of the node name   Why?  H One only needs non-zero allocation classes in two cases: disks connectedE to more than one machine (so that the disk name is independent of the G node name) and if shadowing is used (this puzzled me for a while, since < shadowing---at least with members connected to more than oneH node---makes dual-ported disks less necessary, but the reason is just to= make the name short enough to fit in a data structure used by  shadowing).   I Since disk names have to be unique within the cluster, do all your disks   have different SCSI addresses?   ------------------------------  % Date: Thu, 10 Jul 2003 18:11:07 +0200 , From: "Reinhard Eigner" <antispam@garnix.de> Subject: Re: Clustering Problem / Message-ID: <bek36o$if7$01$1@news.t-online.com>   L "Phillip Helbig---remove CLOTHES to reply" <helbig@astro.multiCLOTHESvax.de>5 schrieb im Newsbeitrag news:bek1t3$c8m$1@online.de... C > In article <bejs5v$cp7$07$1@news.t-online.com>, "Reinhard Eigner"  > <antispam@garnix.de> writes: > 7 > > > GARYV1$DKA100: Mounted 0 OVMSVAXSYS 1399815 222 1  > > > GAZVX2$DKA0: Online 0  > > > GAZVX2$DKA300: Online 0  > > > GAZVX2$DKA400: Online 0  > > @ > > Didn't you have to change the allocation class of the disks?; > > On our ES40 3-node cluster all the disks beginning with & > > $1$DKAxxx instead of the node name >  > Why? > J > One only needs non-zero allocation classes in two cases: disks connectedG > to more than one machine (so that the disk name is independent of the I > node name) and if shadowing is used (this puzzled me for a while, since > > shadowing---at least with members connected to more than oneJ > node---makes dual-ported disks less necessary, but the reason is just to? > make the name short enough to fit in a data structure used by 
 > shadowing).  > J > Since disk names have to be unique within the cluster, do all your disks  > have different SCSI addresses?  + I didn't read this thread correctly, sorry! E Our cluster is connected to a HSZ80-Controller, full of disks and one  cache battery :-)    bye    ------------------------------    Date: 10 Jul 2003 01:51:51 -0700% From: Bart.Zorn@xs4all.nl (Bart Zorn)  Subject: Re: CSWS v1.3 abort= Message-ID: <a98cd882.0307100051.24067aec@posting.google.com>   g "old account" <rwolff@selkirk.bc.ca> wrote in message news:<000501c346a3$9d0bddc0$41933cc7@cassb20c>... A > I upgraded from OpenVMS 7.2-1 to 7.2.2 so I could run CSWS 1.3. B > After installing 7.2-2 plus the TCPIP fix I upgraded to CSWS1.3./ > When I started CSWS I received the following:  >  > $ @sys$startup:apache$startup B > %APACHE-S-PROC_ID, identification of created process is 00000429E > %APACHE-E-PROCTERMINATED, process terminated with status = 1001829A  > %SYSTEM-F-ABORT, abort > %SYSTEM-F-ABORT, abort > $  > ) > Any help would be appreciated.  Thanks.   @ The error code means: %RMS-E-PRV, insufficient privilege or file protection violation  ' Which account do you use to start CSWS?   	 Bart Zorn    ------------------------------  # Date: Thu, 10 Jul 2003 07:13:29 GMT 6 From: peter@langstoeger.at (Peter 'EPLAN' LANGSTOEGER) Subject: Re: CSWS v1.3 abort3 Message-ID: <tm8Pa.40001$1F6.603894@news.chello.at>   b In article <000501c346a3$9d0bddc0$41933cc7@cassb20c>, "old account" <rwolff@selkirk.bc.ca> writes:@ >I upgraded from OpenVMS 7.2-1 to 7.2.2 so I could run CSWS 1.3.A >After installing 7.2-2 plus the TCPIP fix I upgraded to CSWS1.3. . >When I started CSWS I received the following: >  >$ @sys$startup:apache$startupA >%APACHE-S-PROC_ID, identification of created process is 00000429 D >%APACHE-E-PROCTERMINATED, process terminated with status = 1001829A >%SYSTEM-F-ABORT, abort  >%SYSTEM-F-ABORT, abort   ( $ write sys$output f$message(%x1001829A)? %RMS-E-PRV, insufficient privilege or file protection violation    --   Peter "EPLAN" LANGSTOEGER % Network and OpenVMS system specialist  E-mail  peter@langstoeger.atF A-1030 VIENNA  AUSTRIA              I'm not a pessimist, I'm a realist   ------------------------------  % Date: Thu, 10 Jul 2003 11:15:49 -0400 , From: "warren sander" <warren.sander@hp.com> Subject: Re: CSWS v1.3 abort, Message-ID: <3f0d883c$1@usenet01.boi.hp.com>  > make sure that your apache account owns all the correct files.  L if you don't shut csws down during the install then a couple of images don't- get the correct ownership (even if you do the ' set the ownership stuff in the config).   I do a dir/own (or dir/sec) and make sure the csws images can get access to ! everything it needs access to ...     2 "Bart Zorn" <Bart.Zorn@xs4all.nl> wrote in message7 news:a98cd882.0307100051.24067aec@posting.google.com... 7 > "old account" <rwolff@selkirk.bc.ca> wrote in message 1 news:<000501c346a3$9d0bddc0$41933cc7@cassb20c>... C > > I upgraded from OpenVMS 7.2-1 to 7.2.2 so I could run CSWS 1.3. D > > After installing 7.2-2 plus the TCPIP fix I upgraded to CSWS1.3.1 > > When I started CSWS I received the following:  > > ! > > $ @sys$startup:apache$startup D > > %APACHE-S-PROC_ID, identification of created process is 00000429G > > %APACHE-E-PROCTERMINATED, process terminated with status = 1001829A  > > %SYSTEM-F-ABORT, abort > > %SYSTEM-F-ABORT, abort > > $  > > + > > Any help would be appreciated.  Thanks.  > B > The error code means: %RMS-E-PRV, insufficient privilege or file > protection violation > ) > Which account do you use to start CSWS?  >  > Bart Zorn    ------------------------------  % Date: Thu, 10 Jul 2003 11:10:33 -0400 / From: "Joe H. Gallagher" <dtrwiz@ix.netcom.com> & Subject: Re: DEClaser 5100 (LN09) help- Message-ID: <3F0D81E6.B8AE9987@ix.netcom.com>    John Laird wrote:  > K > A client has just had one of these delivered from a reseller.  After much M > fun with the comms, I got it fully working, dcps'd etc.  Just about to pack M > my bags and depart and someone asked what I was going to do "with the other M > bits", lurking under a desk ;-)  Looks like they also have a large capacity % > tray that goes underneath the unit.  > H > I did not have time to start fiddling, but if anyone can provide briefL > instructions on how to attach the large tray, I would be very grateful.  AM > printer user guide was included, which I also found online, but it does not L > go into any detail about options (other than advising on how to adjust theD > spring tension) but I have not been able to track down any optionsN > installation manual or suchlike.  The large tray appears to be complete, andN > clearly a small 4-pin plug connects to a socket on the underside of the mainD > unit.  The menus would seem to suggest the printer can feed from aL > multiplicity of sources, so presumably we can simply drop the printer onto;                                                        ^^^^ < Drop?  No, very carefully place.  If you don't have a lot of< upper body strength, I would recommend two people.  The 5100> is 41 pounds and it doesn't have any handles.  Don't mash your> finders.  Be very careful not to break or bend the 4-pin plug.> You may need to adjust the spring tension to the paper to feed> without jamming.  If the LCIT (large capacity input tray) has = had lots of use, you may have to replace the pickup roller.    It is trivial to do.  = Set the option in the menu to take paper from the LCIT (large ; capacity input tray) first and then take from the standard  = (universal) input try.  That's all there is to it.  You will  ; need to leave some paper in the universal tray as it draws  0 paper from this tray  for the startup report(s).  L > the tray, power it up, select the large tray as the default feed and leaveJ > it like that ?  (This is probably easier to manage for the users than me" > trying to get clever with DCPS.) > K > I did not manage to decipher the odd 3-figure diagram to the left side of  > the large tray.. >  > TIA.   Joe H. Gallagher4 (we self-maintain DECLaser 2150s and DEClaser 5100s) dtrwiz at ix dot netcom dot com    ------------------------------  % Date: Thu, 10 Jul 2003 17:37:19 +0100 + From: John Laird <john@laird-towers.org.uk> & Subject: Re: DEClaser 5100 (LN09) help8 Message-ID: <hg5rgvs795sjftcor9sm1ps9lb9lovlubd@4ax.com>  6 On Thu, 10 Jul 2003 11:10:33 -0400, "Joe H. Gallagher" <dtrwiz@ix.netcom.com> wrote:    >(help as requested)  . Thanks very much.  Reply via email on its way.   --   John   ------------------------------    Date: 10 Jul 2003 10:11:50 -0700/ From: kenneth.randell@verizon.net (Ken Randell) ( Subject: Re: Good code profilers on VMS?= Message-ID: <79de9693.0307100911.180ebd0a@posting.google.com>    > N > > It beats not having one, which is where serious developers are on OpenVMS. >   B Would DCPI (as discussed in the current issue of the VMS TechnicalD journal) be of use here?  I've not used it (on VMS), but it's better than 'not having one'.   ------------------------------  # Date: Thu, 10 Jul 2003 17:26:13 GMT 9 From: "Fred Kleinsorge" <my-last-name@stardotzko.dec.com> 4 Subject: Re: HP World: Why Alpha's Omega Makes Sense2 Message-ID: <VkhPa.4333$yV2.3008@news.cpqcorp.net>  5 "Bill Todd" <billtodd@metrocast.net> wrote in message , news:II-dncxx59R3FpeiXTWJlg@metrocast.net...   <delete>  I Problem here Bill is that while there are some nit's in Terry's story, it L pretty much agrees with most of the *reliable* information I've heard.  YourH "analysis" really comes from building a house of cards based informationJ which is a collection half truths, guesses, rumors, complete fabrications,K and an occasional truth -- and I'm not even saying that *you* are making it K up.  Even if I were to give you some shred of regard for your objectiveness L (and frankly you froth a bit every time you talk about Alpha, Compaq, HP, orG VMS) - your opinions are pretty much based on little-to-no actual real,   insider, or factual information.  D But hey, it fits your world view.  How about trying a little thoughtD experiment.  Go back and re-read it without all your assumptions andL opinions - as if all the facts in the story are indeed verifyable facts.  IsL the story plausable?  Now.  Assume that 80% of everything in both your fairyL tale, and Terry's fairy tale are shaded by someone having had an agenda whenL they provided the factual basis of each story.  Which one is more plausable?  K Now.  Really suck it up guy.  Realize that this isn't an episode on Dallas, H and you aren't going to wake up and find that it was all a dream.  AlphaJ will fade away.  IA64 has achieved at least parity in performance with theK leading 64 bit ISA's (we can argue fine details in the small percentages of K who has the absolute lead in what specific area).  IA64 has a number of big E supporters, runs a number of useful OS's, runs on the entire range of J systems from 1 CPU to big/crazy sized - from *real* tier 1 companies - andK has a growing list of ISV's.  It's now 2 years later, and OpenVMS has still J not been cancelled.  Internally we are getting tremendous support from theG platform hardware groups, the architecture groups, the chip groups, the L HP-UX and Linux groups and we *are* on all our internal roadmaps with no endF date, and we are now a peer organization with HP-UX and the HW groups.  L OK.  Well, I guess you can stick to your "we're all dirty liars out to screwC customers and we can't be trusted" (I believe you want some type of I purification ritual that includes the sacrifice of a CEO before we can be L rehabilitated).  In fact, pre-merger Compaq was upfront with customers aboutK the future at the same time as they told us (VMS) about the new direction - I even if it took some time and distance to know the underlying reasons for I the change and how it came about.  Maybe they played it wrong, maybe not. G But for the last 2 years we have been doing pretty much what we said we K would do, and VMS is lined up to deliver IA64 system support pretty much as I it becomes the fastest/biggest/best 64-bit platform.  HP has embraced VMS  from the bottom up.   K In 2004, OpenVMS customers will be running VAX's, Alpha's, and IPF systems. J They'll still be buying Alpha's.  Architecture migrations are a multi-yearL thing.  Mixed clusters and Alpha support will be around a *long* time - lookI at the antique VAXes we still are supporting.  We're still supporting the J DEC 3000 Alpha's - *years* after Tru64 dropped support for them.  I expectJ IPF to simply get bigger/smaller/better/faster/cheaper, and yes - I expectJ that there will only be a couple 64 bit architectures that survive - and IH would not bet on Opteron, or Sparc to be among them - and definately notL both.  And I expect in 2010 there will still be Bill Todd predictions of theJ end of the world as we know it, and I also expect that we will probably be4 on OpenVMS for Itanium and Alpha - V10.x by then ;-)   Chill Bill.    ------------------------------  % Date: Thu, 10 Jul 2003 17:59:32 +0100 / From: Tim ffrench-Lynch <nospam@baesystems.com>  Subject: Re: KVM switches . Message-ID: <3F0D9B74.C77DAB10@baesystems.com>   Ken Fairfield wrote:@ > Anyone out there actively using the Adder GEM with an Alpha???  D I should be receiving an Adder Omega 2 port KVM on Monday. I'll post8 here how well it performs with an Alphastation and a PC.   Tim    ------------------------------  % Date: Thu, 10 Jul 2003 17:45:16 +0300 " From: Guy Peleg <guy.peleg@hp.com>* Subject: New lexical function F$DELTA_TIME& Message-ID: <3F0D7BFC.383CB534@hp.com>   Hello DCL gurus,  G I thought you might be interested  to know that with the next Alpha VMS " version V7.3-2, we will ship a new lexical function F$DELTA_TIME.  D F$DELTA returns the difference between a beginning and end time. The# first argument is an absolute time, D the second argument can be absolute or delta time. See the following example.   BLUSKY> start=f$time() BLUSKY> end=f$time() BLUSKY> sh sym start#   START = "10-JUL-2003 17:38:06.82"  BLUSKY> sh sym end!   END = "10-JUL-2003 17:38:30.64" + BLUSKY> write sys$output f$delta(start,end)     0 00:00:23.82  ( No it won't be backported to VAX 5.5-2 .  $ As usual, your comments are welcome.  
 Guy Peleg, OpenVMS Engineering    ------------------------------  + Date: Thu, 10 Jul 2003 15:26:50 +0000 (UTC) - From: lewis@spyder.mitre.org (Keith A. Lewis) . Subject: Re: New lexical function F$DELTA_TIME. Message-ID: <bek0jq$ltq$1@newslocal.mitre.org>  p Guy Peleg <guy.peleg@hp.com> writes in article <3F0D7BFC.383CB534@hp.com> dated Thu, 10 Jul 2003 17:45:16 +0300:H >I thought you might be interested  to know that with the next Alpha VMS# >version V7.3-2, we will ship a new  >lexical function F$DELTA_TIME.  > E >F$DELTA returns the difference between a beginning and end time. The $ >first argument is an absolute time,E >the second argument can be absolute or delta time. See the following 	 >example.  >  >BLUSKY> start=f$time()  >BLUSKY> end=f$time()  >BLUSKY> sh sym start $ >  START = "10-JUL-2003 17:38:06.82" >BLUSKY> sh sym endP" >  END = "10-JUL-2003 17:38:30.64", >BLUSKY> write sys$output f$delta(start,end) >   0 00:00:23.82m    That's a useful addition, Guy.    I But there's still one thing missing from DCL date/time lexicals (or maybebK just from my knowledge of them).  Suppose I want to add or subtract a deltalI time from an absolute time.  With a delta of up to 23:59:59.99 I can use P  5     f$cvtime("''base_time'+''delta_time'","ABSOLUTE") 5     f$cvtime("''base_time'-''delta_time'","ABSOLUTE")c  G But this only works for delta_time < 24 hours.  In system services it's G easy to add and subtract times up to 9999 days.  Can it be done in DCL?e  + --Keith Lewis              klewis$mitre.orgu> The above may not (yet) represent the opinions of my employer.   ------------------------------   Date: 10 JUL 2003 15:35:29 GMT+ From: Dave Greenwood <greenwoodde@ornl.gov>e. Subject: Re: New lexical function F$DELTA_TIME2 Message-ID: <10JUL03.15352971@feda34.fed.ornl.gov>  : In a previous article, Guy Peleg <guy.peleg@hp.com> wrote: > Hello DCL gurus, >  eI > I thought you might be interested  to know that with the next Alpha VMSa$ > version V7.3-2, we will ship a new  > lexical function F$DELTA_TIME. >  wF > F$DELTA returns the difference between a beginning and end time. The% > first argument is an absolute time,fF > the second argument can be absolute or delta time. See the following
 > example. >  ] > BLUSKY> start=f$time() > BLUSKY> end=f$time() > BLUSKY> sh sym start% >   START = "10-JUL-2003 17:38:06.82"c > BLUSKY> sh sym end# >   END = "10-JUL-2003 17:38:30.64"-- > BLUSKY> write sys$output f$delta(start,end)u >    0 00:00:23.82  F I like it.  Presumably the difference between the arguments is limitedH to 9999 days?  Do you get an absolute time result if the second argument is a delta time?  * > No it won't be backported to VAX 5.5-2 .  
 Awwwww. ;)   Dave --------------9 Dave Greenwood                Email: Greenwoodde@ORNL.GOVsH Oak Ridge National Lab        %STD-W-DISCLAIMER, I only speak for myself   ------------------------------  # Date: Thu, 10 Jul 2003 15:40:54 GMTr" From:   VAXman-  @SendSpamHere.ORG. Subject: Re: New lexical function F$DELTA_TIME0 Message-ID: <00A22A60.98A2B4E2@SendSpamHere.ORG>  K In article <3F0D7BFC.383CB534@hp.com>, Guy Peleg <guy.peleg@hp.com> writes:- >Hello DCL gurus,1
 >{...snip...}9  4 Guy, do you have an email address outside of hp.com?  F I'd like to ask you a few questions and make a request concerning DCL.   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COM              5   "Well my son, life is like a beanstalk, isn't it?" l   ------------------------------  + Date: Thu, 10 Jul 2003 15:50:11 +0000 (UTC)s+ From: david20@alpha1.mdx.ac.uk (David Webb)h. Subject: Re: New lexical function F$DELTA_TIME+ Message-ID: <bek1vj$c0e$1@aquila.mdx.ac.uk>e  ^ In article <bek0jq$ltq$1@newslocal.mitre.org>, lewis@spyder.mitre.org (Keith A. Lewis) writes:q >Guy Peleg <guy.peleg@hp.com> writes in article <3F0D7BFC.383CB534@hp.com> dated Thu, 10 Jul 2003 17:45:16 +0300:nI >>I thought you might be interested  to know that with the next Alpha VMSh$ >>version V7.3-2, we will ship a new  >>lexical function F$DELTA_TIME. >>F >>F$DELTA returns the difference between a beginning and end time. The% >>first argument is an absolute time,rF >>the second argument can be absolute or delta time. See the following
 >>example. >> >>BLUSKY> start=f$time() >>BLUSKY> end=f$time() >>BLUSKY> sh sym start% >>  START = "10-JUL-2003 17:38:06.82"  >>BLUSKY> sh sym end# >>  END = "10-JUL-2003 17:38:30.64"b- >>BLUSKY> write sys$output f$delta(start,end)h >>   0 00:00:23.82 >e! >That's a useful addition, Guy.    >sJ >But there's still one thing missing from DCL date/time lexicals (or maybeL >just from my knowledge of them).  Suppose I want to add or subtract a deltaJ >time from an absolute time.  With a delta of up to 23:59:59.99 I can use  >16 >    f$cvtime("''base_time'+''delta_time'","ABSOLUTE")6 >    f$cvtime("''base_time'-''delta_time'","ABSOLUTE") >IH >But this only works for delta_time < 24 hours.  In system services it'sH >easy to add and subtract times up to 9999 days.  Can it be done in DCL? >C   You mean like :-   Alpha2:sh time   10-JUL-2003 16:41:598 Alpha2:a = f$cvtime("''f$time()' +6-4:0:0.0","ABSOLUTE") Alpha2:sh sym a4   A = "16-JUL-2003 20:42:05.84"9   and0   Alpha2:sh time   10-JUL-2003 16:44:038 Alpha2:a = f$cvtime("''f$time()' -5-4:0:0.0","ABSOLUTE") Alpha2:sh sym a:   A = "5-JUL-2003 12:44:06.66"  N Note there cannot be any space between the + (or minus) and the number of days
 else you get 5  9 Alpha2:a = f$cvtime("''f$time()' + 6-4:0:0.0","ABSOLUTE")1J %DCL-W-IVATIME, invalid absolute time - use DD-MMM-YYYY:HH:MM:SS.CC format&  \10-JUL-2003 16:46:11.98 + 6-4:0:0.0\  
 David Webb VMS and Unix team leader CCSS Middlesex University      , >--Keith Lewis              klewis$mitre.org? >The above may not (yet) represent the opinions of my employer.0   ------------------------------   Date: 10 Jul 2003 16:21:21 GMT1 From: JONESD@er6.eng.ohio-state.edu (David Jones)n. Subject: Re: New lexical function F$DELTA_TIME: Message-ID: <bek3q1$qlh$1@charm.magnus.acs.ohio-state.edu>   Speaking of lexicals...O  N What I'd like is for f$parse("$3$dka100:[foo]bar.dat",,,"NAME+TYPE") to return
 "bar.dat".    < David L. Jones               |      Phone:    (614) 292-6929- Ohio State University        |      Internet:nL 140 W. 19th St. Rm. 231a     |               jonesd@er6s1.eng.ohio-state.edu: Columbus, OH 43210           |               vman+@osu.edu  1 Disclaimer: I'm looking for marbles all day long.o   ------------------------------  % Date: Thu, 10 Jul 2003 19:47:14 +03000" From: Guy Peleg <guy.peleg@hp.com>. Subject: Re: New lexical function F$DELTA_TIME& Message-ID: <3F0D9891.A580E1BF@hp.com>   The only one I have are @ guy.peleg@hp.com, guy.peleg@digital.com and guy.peleg@compaq.com peleg@star.zko.dec.com   What's wrong with HP mail?  ! VAXman-, @SendSpamHere.ORG wrote:r  M > In article <3F0D7BFC.383CB534@hp.com>, Guy Peleg <guy.peleg@hp.com> writes:k > >Hello DCL gurus,o > >{...snip...}A >S6 > Guy, do you have an email address outside of hp.com? >tH > I'd like to ask you a few questions and make a request concerning DCL. >n > --Q > VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COM4 >x6 >   "Well my son, life is like a beanstalk, isn't it?"   ------------------------------    Date: 10 Jul 2003 10:51:33 -0700, From: JimStrehlow@data911.com (Jim Strehlow). Subject: Re: New lexical function F$DELTA_TIME< Message-ID: <4b6ec350.0307100951.532460a@posting.google.com>  P Guy Peleg <guy.peleg@hp.com> wrote in message news:<3F0D7BFC.383CB534@hp.com>...> ... with the next Alpha VMS version V7.3-2, we will ship a new! lexical function F$DELTA_TIME ...r     Having optional parameters for4 (..., "day"|"hour"|"minute"|"second", variableName ): would be "nice" to save us from parsing the result string;% but we can parse if we have to do so.s  T It would save us from writing an extra couple of lines of code (our own subroutine).   Jim, Data911, Alameda, CA, USA   ------------------------------  % Date: Thu, 10 Jul 2003 12:55:45 +0530n From: "Vivek" <visoni@bmc.com> Subject: Old Mailbox Problem/ Message-ID: <vgq51ubfm8uv6c@corp.supernews.com>e   Hi Mezei, Forrest and all,  J The problems which I raised about the inefficiency of the mailbox is over.  J The problem was not with the Mailbox but other spurious calculations being% done before writing data in the code.cK So there was not data in the Mailbox and the read was waiting for the same.   J The Mailboxes could be written to and read out from in microseconds. So we& dont have to doubt them on this later.   Thanks for all the Helpu Vivekz   ------------------------------    Date: 10 Jul 2003 07:18:39 -0500; From: koehler@eisner.nospam.encompasserve.org (Bob Koehler):$ Subject: Re: OpenVMS I64, a proposal3 Message-ID: <9pGEwmAmpnbU@eisner.encompasserve.org>   c In article <5XzHgOLm8I6L@eisner.encompasserve.org>, Kilgallen@SpamCop.net (Larry Kilgallen) writes:t > A > I hate to consider all that freeware floating around built with,? > still-buggy compilers.  Presumably once IA64 VMS is released, ? > anything that built on Alpha VMS should easily build, or elsenD > depends on IA64 VMS internals that _do_ require listing kits, etc.  C    When I got my first Alpha, running OpenVMS Alpha AXP 1.5, there oG    were bugs in the Fortran and C compilers I hadn't expected.  Didn't -G    have a big impact on porting all my stuff from VAXen I clustered it E    with.  D    But then I was lucky, since I learned "VAX-11 Macro" writing codeH    that needed to interface with Fortran my Macro-32 code I didn't have H    any JSB,BSB,RSB,... and that needed help on ALpha.  My Macro-32 code     ported 100% untouched.n    r   ------------------------------  % Date: Thu, 10 Jul 2003 06:52:23 -04001) From: "Neil Rieck" <n.rieck@sympatico.ca>t8 Subject: Re: OpenVMS Technical Seminar Highlights (some)9 Message-ID: <IzbPa.9276$Ag6.586661@news20.bellglobal.com>i  I While re-reading my seminar notes from the DECUS Canada OpenVMS Technical-M Seminar, I remembered something a little odd: at various times they would ask0M for a show of hands for  questions like "who uses CSWS?", "who is still usingeK VAX-only clusters?", "who backs up their systems with $BACKUP" etc. The oddgN bit came when they asked "Who uses RDB?" and more that 2/3 (almost 3/4) of theJ people raised their hands but the HP people didn't seem surprised and just nodded.b  L Maybe I'm reading too much into this, but I wonder if RDB will be moved fromH Oracle back to OpenVMS engineering? (Although this might not be fiscally4 possible unless Larry Ellison is feeling charitable)  
 Neil Rieck Kitchener/Waterloo/Cambridge,d Ontario, Canada.! http://www3.sympatico.ca/n.rieck/    ------------------------------    Date: 10 Jul 2003 06:52:23 -0500- From: Kilgallen@SpamCop.net (Larry Kilgallen)r8 Subject: Re: OpenVMS Technical Seminar Highlights (some)3 Message-ID: <1RBQm4AkvzWU@eisner.encompasserve.org>   e In article <IzbPa.9276$Ag6.586661@news20.bellglobal.com>, "Neil Rieck" <n.rieck@sympatico.ca> writes: K > While re-reading my seminar notes from the DECUS Canada OpenVMS TechnicaloO > Seminar, I remembered something a little odd: at various times they would ask O > for a show of hands for  questions like "who uses CSWS?", "who is still usingaM > VAX-only clusters?", "who backs up their systems with $BACKUP" etc. The oddcP > bit came when they asked "Who uses RDB?" and more that 2/3 (almost 3/4) of theL > people raised their hands but the HP people didn't seem surprised and just	 > nodded.  > N > Maybe I'm reading too much into this, but I wonder if RDB will be moved fromJ > Oracle back to OpenVMS engineering? (Although this might not be fiscally6 > possible unless Larry Ellison is feeling charitable)  F If Rdb were an HP product, when Oracle was asked what operating systemF to use for Classic Oracle they would have a financial incentive to sayH "anything but VMS", since VMS would be the platform where Classic Oracle had strong competition.M   ------------------------------  # Date: Thu, 10 Jul 2003 12:39:26 GMT0# From: "John Smith" <a@nonymous.com>68 Subject: Re: OpenVMS Technical Seminar Highlights (some)I Message-ID: <28dPa.119259$x4o.47788@news04.bloor.is.net.cable.rogers.com>   : "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message- news:1RBQm4AkvzWU@eisner.encompasserve.org...0A > In article <IzbPa.9276$Ag6.586661@news20.bellglobal.com>, "Neil8% Rieck" <n.rieck@sympatico.ca> writes:4C > > While re-reading my seminar notes from the DECUS Canada OpenVMS 	 Technical0B > > Seminar, I remembered something a little odd: at various times they would askE > > for a show of hands for  questions like "who uses CSWS?", "who is0 still usingVB > > VAX-only clusters?", "who backs up their systems with $BACKUP" etc. The oddF > > bit came when they asked "Who uses RDB?" and more that 2/3 (almost 3/4) of thePE > > people raised their hands but the HP people didn't seem surprised: and just > > nodded.n > >7E > > Maybe I'm reading too much into this, but I wonder if RDB will bec
 moved fromC > > Oracle back to OpenVMS engineering? (Although this might not be0 fiscally8 > > possible unless Larry Ellison is feeling charitable) > A > If Rdb were an HP product, when Oracle was asked what operatingC systemD > to use for Classic Oracle they would have a financial incentive to say C > "anything but VMS", since VMS would be the platform where ClassicS Oracle > had strong competition.   F At this stage of non-advertising/marketing of VMS, handing Rdb back toD HP would give Oracle plenty of incentive to drop Oracle 9(etc..) andE successors on VMS, much in the same way as Sybase EOL'd their product1@ on VMS. As pointed out, Rdb on VMS is a formidable competitor to1 Oracle 'Classic', if it were advertised/marketed.1  B Larry wouldn't care that VMS wasn't supported, only that customers? stayed with Oracle on Unix or Linux. And given the way the vastCC majority of dbms users lock themselves into their dbms tighter than0F they lock themselves into the o/s, if Oracle Classic were dropped from6 VMS, those customers would tend to move to unix/linux.   ------------------------------  + Date: Thu, 10 Jul 2003 12:53:09 +0000 (UTC)1P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply)8 Subject: Re: OpenVMS Technical Seminar Highlights (some)$ Message-ID: <bejnjl$41p$2@online.de>  F In article <IzbPa.9276$Ag6.586661@news20.bellglobal.com>, "Neil Rieck" <n.rieck@sympatico.ca> writes:    N > Maybe I'm reading too much into this, but I wonder if RDB will be moved fromJ > Oracle back to OpenVMS engineering? (Although this might not be fiscally6 > possible unless Larry Ellison is feeling charitable)  F Maybe this was one sell-off which DID make sense (though I would have E asked for more than one year's revenue): If Rdb is with DEC, then it mI competes with Oracle, so Oracle will have less interest promoting Oracle uF "Classic" on VMS.  It's been a while since Rdb went to Oracle, but it I still seems very "VMS-like" to me.  (This assumes that there is value in  ( having Oracle Classic on VMS.  Discuss.)   ------------------------------    Date: 10 Jul 2003 09:34:13 -0500- From: Kilgallen@SpamCop.net (Larry Kilgallen)C8 Subject: Re: OpenVMS Technical Seminar Highlights (some)3 Message-ID: <SeQtGKDl2wkA@eisner.encompasserve.org>@  w In article <bejnjl$41p$2@online.de>, helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply) writes:mH > In article <IzbPa.9276$Ag6.586661@news20.bellglobal.com>, "Neil Rieck"! > <n.rieck@sympatico.ca> writes:   > O >> Maybe I'm reading too much into this, but I wonder if RDB will be moved fromeK >> Oracle back to OpenVMS engineering? (Although this might not be fiscallye7 >> possible unless Larry Ellison is feeling charitable)4 > H > Maybe this was one sell-off which DID make sense (though I would have G > asked for more than one year's revenue): If Rdb is with DEC, then it oK > competes with Oracle, so Oracle will have less interest promoting Oracle  H > "Classic" on VMS.  It's been a while since Rdb went to Oracle, but it K > still seems very "VMS-like" to me.  (This assumes that there is value in l* > having Oracle Classic on VMS.  Discuss.)  C There is value in having such a prominent database vendor as Oraclec support VMS.   ------------------------------  % Date: Thu, 10 Jul 2003 09:56:27 +0200e$ From: "Peter Flunger" <p-i-b@gmx.at>= Subject: Re: Opinions please: EVA and database journal policy 0 Message-ID: <bej67d$cli$1@newsreader1.netway.at>  / "Jeff Goodwin" <jgoodwin@maine.rrr-r.com> wroteeJ > Storage allocation on the EVA makes this practice less clear.  I guess IK > want to know if people are allocating separate storage pools for database I > and journals or if one storage pool is sufficient.  If I'm missing somer* > other option, I'd like to hear that too. >jJ In one our sites we are using two disk groups in every EVA, one for Oracle" data ( Tablespaces, Redologfiles )J and the other one for the rest ( including archivelog files, comparable to RDBs AIJ files ) plus@ we let Oracle mirror the online-redologfiles two one unit in the8 Oracle diskgroup and one unit in the 'other' disk group. plusF everything is shadowed to another EVA in another site using HBVS (host based volume shadowing)nD We cannot afford even the shortest down times, but on the other hand maybe we are just paranoid :)a Petere   ------------------------------  # Date: Thu, 10 Jul 2003 15:54:44 GMTl- From: "Kelvin Smith" <fcs_smith1111@snet.net>b$ Subject: Re: PDP-11 OS Release DatesA Message-ID: <8%fPa.9084$Ng.2661065815@newssvr10.news.prodigy.com>a  L I threw out most of my old RSTS information when the office moved last year, but I have SPD's for v10:A   RSTS v10.0, July 1990u RSTS v10.1, Sep 1992  9 I got started with RSTS in 1976, using either v6A or v6B.    -- Kelvin Smith Financial Computer Systems
 Wilton, CT& (remove "1111" from address for email)  = "Zane H. Healy" <healyzh@shell1.aracnet.com> wrote in messageH% news:behgbr047t@enews4.newsguy.com...rK > For a start, I'd like to appologize for the cross-posting, but I'm tryinge toF > reach people that might have info but don't normally read the PDP-11 groups.A >:I > I'm looking for release dates for the various PDP-11 OS's.  For example2 for0 > RT-11:   ------------------------------  # Date: Thu, 10 Jul 2003 17:09:31 GMTo' From: Don Sykes <anonymous@pacbell.net>  Subject: Pearl Users+ Message-ID: <3F0D9DFC.8CB8FA99@pacbell.net>f  C I don't know Pearl at all, but it seems to be very popular with COV  folks.; Can anyone give me a brief list of its advantages over DCL?k   TIAs -- p   Have VMS, Will Travel) Wire paladin, San Franciscoi   (paladinATalphaseDOTcom)   ------------------------------  % Date: Thu, 10 Jul 2003 13:37:35 -0400m8 From: Jim Agnew - VCU/MCV Neurosurgery <jpagnew@vcu.edu> Subject: Re: Pearl Users' Message-ID: <3F0DA45F.38CABAE2@vcu.edu>b  = portablity...  you can write a perl program, and then take it F elsewhere.  like for example, you can do some kind of database report,D using the perl dbi/dbd modules to create a seamless call to most anyE database supported by the modules, making operating system, database,e
 irrelevant...   H little gotcha's exist, tho, just like on most anything else, like one db9 likes " in the sql syntax, db2 likes ', things like that.i  H otherwise, it one heck of a language, and it's no real slouch with speed either.>  F you *do have a steep learning curve, but who says you have to know allH of the language to be able to do simple data editing scripts in it, thatH you can batch off insead of doing it interactively with edit/edt or /tpu or /teco...   ! it's almost as fast as TECO, too.    jim    Don Sykes wrote: > E > I don't know Pearl at all, but it seems to be very popular with COV  > folks.= > Can anyone give me a brief list of its advantages over DCL?  >  > TIAs > -- >  > Have VMS, Will Traveli > Wire paladin, San Franciscod >  > (paladinATalphaseDOTcom)   -- hF "4,000 years ago I made a mistake."  Elrond Half-Elven, in "Fellowship of the Ring"  F "I try not to be right any more than necessary". -- Larry Wall, author of the Perl Language   ------------------------------  % Date: Thu, 10 Jul 2003 10:08:54 +0100 " From: Didier Morandi <no@spam.com> Subject: PPPOE? When? How?' Message-ID: <3F0D2D26.7070208@spam.com>d   Hello happy tax payers,   ( I want to pppoe from my Alpha to my ISP.J I found in the FAQ that there is no such PPP for the moment for obscure=20 reasons (to me).  6 Did anyone succeed to do that with a particular tool ?   Thanks,1   D. --=20 - Didier Morandi sarl au capital de 8 000 eurosC                     Tout VMS-   19 chemin de la Butte 31400 Toulouse France /   T=E9l: 33(0)5 6120 1964 Fax: 33(0)5 6154 1928 &           http://www.didiermorandi.com$             RCS Toulouse 448 694 851   ------------------------------  % Date: Thu, 10 Jul 2003 16:15:33 +0200p, From: "Reinhard Eigner" <antispam@garnix.de> Subject: Re: PPPOE? When? How?/ Message-ID: <bejse3$q86$05$1@news.t-online.com>   , > > I want to pppoe from my Alpha to my ISP.K > > I found in the FAQ that there is no such PPP for the moment for obscuret > > reasons (to me). > >-: > > Did anyone succeed to do that with a particular tool ? >u< > Yes, with an integrated modem-router (on a DSL connexion). > = I suggest the same. I also use a DSL router at home, where myc4 two PC Systems and my PWS600au@VMS7.3-1 are running.< That's the easiest way to bring your machine in the internet   ------------------------------  + Date: Thu, 10 Jul 2003 12:47:02 +0000 (UTC)1P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply)/ Subject: Re: problem with MONITOR CLUSTER (VPM)t$ Message-ID: <bejn86$41p$1@online.de>  J > I have seen this a few times.. just stop the VPM Server processes on allJ > nodes and try your Monitor command again... I have not seen a difinitiveG > answer on how to keep from scrambling the VPM Server's brains or whatO > causes it.  + Killed them all, restarted, same behaviour.E   ------------------------------    Date: 10 Jul 2003 06:32:38 -0700( From: robert_kersey@bat.com (Rob Kersey)  Subject: Re: Restoring an backup= Message-ID: <f936a854.0307100532.47997910@posting.google.com>o   Try using /select   	 To backup   0 Back/Log [...] /select=[...]*.*;* savset.bck/sav   Restoree  0 Back/Log Saveset.bck/sav/select=[...]*.*;* [...]   ------------------------------  # Date: Thu, 10 Jul 2003 16:18:49 GMT 9 From: "Fred Kleinsorge" <my-last-name@stardotzko.dec.com>d Subject: Re: Rethinking  V.M.S2 Message-ID: <JlgPa.4318$CR2.2723@news.cpqcorp.net>  H "Bob Koehler" <koehler@eisner.nospam.encompasserve.org> wrote in message- news:fuKOVVLN7skR@eisner.encompasserve.org...t >iG >    OBTW, some CPU's, like MIPS, simply don't map some addresses.  ForyI >    those addresses the virtual address must be the same as the physicalrG >    address.  This simplifies some OS kernel issues, but I don't thinkd* >    it provides a real performance boost. >4  C Most UNIXes use modes which allow them to map their kernel in a 1:1gH virtual/physical address space (some upper bits in the VA put it in thatE address range).  It's just one of the reasons that most UNIXes have anJ generic installation kernel, and you build your kernel for the platform itK will run on.  It does provide some performance and simplicity advantages toeL have your entire kernel not taking any TLB misses.  You can work around thatK (and VMS does to some degree) with other strategies (like using granularity : hints, or huge pages when loading kernel images and data).   ------------------------------  # Date: Thu, 10 Jul 2003 16:42:54 GMTb- From: Mike Rechtman <michael.rechtman@hp.com>m Subject: Re: Running VMS off CDi& Message-ID: <3F0DBB2E.7EBF7D48@hp.com>   Doc.Cypher wrote:o > N > For those that have noticed I have an interest in remailers (duh!). The goodM > news is that the new Type-III remailer (Mixminion) is written in Python andc# > will therefor be runnable on VMS.9 > L > Any system such as that should be as secure as possible, and I'm wondering6 > about the possibility of booting and running off CD. > N > Has anyone done this?  If so, is XFC enough of a performance booster to makeM > this a viable proposition?  What would be a reasonable amount of memory for 
 > the system?  >  > Thanks for any advice. >  > Doc. > --M > OpenVMS.         Eight out of ten hackers prefer *other* operating systems.   ? Did that recently, more or less as an excercise, mostly on VAX:i: 1. On Alpha set WLKSYSDSK to 1 (*before* you burn your CD)E 2. On VAX I edited a few SYS$STARTUP:VMS*_50-*.com files, to cut log-o    file requirements. H 3. Still on VAX - During startup I see an error relating to the security     database, but boot continued.@ 4. If you have sufficient memory, create a small RAMdisk. E.g. -
 ApparentlyH    SYSUAF.DAT has to be writable to allow user logins. Putting a swap or<    page-file on a RAMdisk would seem to be a waste though...  C My starting point was an image-copied  V7.3-1 distro to an LD disk,i	 (V7.3 forr VAX) and worked from there   Mike.  -- rE --------------------------------------------------------------------- E Usual disclaimer: All opinions are mine alone, perhaps not even that.o? Mike Rechtman                            *rechtman@tzora.co.il*sF Kibbutz Tzor'a.                          Voice (home): 972-2-9908337  B   "20% of a job takes 80% of the time, the rest takes another 80%"E ---------------------------------------------------------------------l -----BEGIN GEEK CODE BLOCK-----a Version: 3.1: GCM/CS d(-)pu s:+>:- a++ C++ U-- L-- W++ N++ K? w--- V+++$6 PS+ PE-- t 5? X- tv-- b+ DI+ D-- G e++ h--- r+++ y+++@ ------END GEEK CODE BLOCK------t   ------------------------------    Date: 10 Jul 2003 03:43:06 -0700" From: daveparboo@hotmail.com (DXP) Subject: System symbolsp= Message-ID: <819f1cc4.0307100243.2a07a9df@posting.google.com>    Greetings Guruso  C In much the same way that it is possible to access "Sys$GL_IJobCnt" B for the interactive job count, what are the System Symbols for CPU Busy/Direct IO/Buffered IO?i   Many tanks in a danceS     Dave   ------------------------------   Date: 10 Jul 03 06:14:12 PST From: mckinneyj@cpva.saic.com  Subject: Re: System symbols ( Message-ID: <JyYrcuFErd9B@cpva.saic.com>  = In article <819f1cc4.0307100243.2a07a9df@posting.google.com>,-%  daveparboo@hotmail.com (DXP) writes:1 > Greetings Gurusm > E > In much the same way that it is possible to access "Sys$GL_IJobCnt" D > for the interactive job count, what are the System Symbols for CPU > Busy/Direct IO/Buffered IO?  >  > Many tanks in a dancex >  >  > Dave     Here's a start -  
 	pms$gl_bufiov
 	pms$gl_dirioy   ------------------------------  # Date: Thu, 10 Jul 2003 13:21:24 GMT " From:   VAXman-  @SendSpamHere.ORG Subject: Re: System symbolse0 Message-ID: <00A22A4D.1C0816CB@SendSpamHere.ORG>  b In article <819f1cc4.0307100243.2a07a9df@posting.google.com>, daveparboo@hotmail.com (DXP) writes: >Greetings Gurus > D >In much the same way that it is possible to access "Sys$GL_IJobCnt"C >for the interactive job count, what are the System Symbols for CPUe >Busy/Direct IO/Buffered IO? >e >Many tanks in a dance >. >  >Dave-  B These values are found in symbolic locations prefixed with PMS$.  * Buffered IO, for example, is PMS$GL_BUFIO.  C There is no cell maintaining "CPU busy" time.  You'd need to sum up-* the times for each of the modes (K,E,S,U).  D Instead of accessing these cells directly, there is a system serviceD (undocumented) SYS$GETSPI to obtain performance information and new-; er versions of VMS contain a documented service SYS$GETRMI.c   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COMo            d5   "Well my son, life is like a beanstalk, isn't it?" e   ------------------------------  + Date: Thu, 10 Jul 2003 10:33:44 +0000 (UTC)tP From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply)D Subject: typos in SYS$COMMON:[SYSMGR]AGEN$NEW_NODE_DEFAULTS.TEMPLATE$ Message-ID: <bejfe8$srp$1@online.de>   ! ; ! This file contains a set of parameter statements specific 6 ! to satellite nodes. The parameter statements in this9 ! file will be automatically read each time MODPARAMS.DAT 7 ! is processed on any non-satellite node added to a VMSh# ! cluster using CLUSTER_CONFIG.COM.s !a5 ! You can edit this file to include any site-specifica8 ! MODPARAMS.DAT statements that are appropriate for your9 ! environment.  Typically you would specify site-specifica: ! minimum or maximum values for certain system parameters. !h  ? Obviously, it should be NON-satellite nodes in the second line!.  8 Also, SYS$COMMON:[SYSMGR]AGEN$NEW_NODE_DEFAULTS.TEMPLATE   says:s   !n9 ! This file contains set of parameter statements specifice6 ! to satellite nodes. The parameter statements in this9 ! file will be automatically read each time MODPARAMS.DATn3 ! is processed on any satellite node added to a VMS # ! cluster using CLUSTER_CONFIG.COM.c !e5 ! You can edit this file to include any site-specifice8 ! MODPARAMS.DAT statements that are appropriate for your9 ! environment.  Typically you would specify site-specific-: ! minimum or maximum values for certain system parameters. !   3 Obviously, the files should be identical except foraA satellite/non-satellite.  So "file contains set" should be "file h contains a set".  H To get back to my thread on SYS$SPECIFIC/SYS$COMMON/SYS$CLUSTER etc, it B seems to me that identical parameter settings in MODPARAMS.DAT on E different nodes in a cluster (EXPECTED_VOTES for example) are a good  I example of something which is identical on each boot disk on the cluster "H but should not be located off the boot disk, so that all the parameters G are available when booting, even if the non-boot disk is not available.r   ------------------------------  # Date: Thu, 10 Jul 2003 16:26:38 GMTt9 From: "Fred Kleinsorge" <my-last-name@stardotzko.dec.com>o> Subject: Re: vms security model - does it still exist on IA64?2 Message-ID: <2tgPa.4319$7S2.4221@news.cpqcorp.net>  6 "bob smith" <sfmc68@bellatlantic.net> wrote in message) news:3F0A0861.2010705@bellatlantic.net...  >  > I > The IA64 does seem to have some hardware capabilities but the area I amtC > interested in is what was changed with ref to the security model.t  J The security model is identical.  The only real changes that I know of areL the lack of a couple obscure combinations of page access modes (combinationsJ that aren't really used, and not in a place that effects user code, or forH that matter any privleged code we can tell).  We also plan on making theF stack no-execute (I believe we are also doing it for Alpha, but it was simpler to do on IA-64).  5 The higher level stuff is all common code with Alpha.    ------------------------------    Date: 10 Jul 2003 11:03:15 -0500+ From: young_r@encompasserve.org (Rob Young)n/ Subject: Re: XFC, IO Size and Backup - Take Two 3 Message-ID: <tK1NSCJOuvIE@eisner.encompasserve.org>T  a In article <4JrWZ6GY$Y46@eisner.encompasserve.org>, young_r@encompasserve.org (Rob Young) writes:n >   ; 	The fun never ends.  For folks following this saga - it is ? 	a journey.  I apologize for not being more thorough initially.3  : 	I realized something else.  I realized you won't get read3 	aheads if XFC is doing large IOs and you have the B; 	default VCC_MAX_IO_SIZE of 127.  Some of what I will writee7 	is speculation - I hope what is speculation is clearlyd 	identifiable.  6 	Taking a 4.8 million block file and backing up to TSM2 	via an ABC client with VCC_MAX_IO_SIZE set to 127 	looks like this:-  * Total QIOs to this file:              9376* Read IOs to this file:                9376* Read Hits:                               0+ Hit Rate:                            0.00 % L Read Aheads:                             0   <<<<<==========================* Read Throughs:                           05 Read Arounds:                         9376 (/NOCACHE)n> Read Arounds:                            0 (Function modifier)4 Read Arounds:                         9376 (IO Size)  : 	By now, we can reverse engineer and determine exactly how 	large that file is:  @ 		512 * 9376 = 4800512 (actual size not quite that big but real 	 		close).    	Raise VCC_MAX_IO_SIZE to 1000:-  * Total QIOs to this file:             18752* Read IOs to this file:               18752* Read Hits:                               0+ Hit Rate:                            0.00 %,H Read Aheads:                          2604   <<<<=======================* Read Throughs:                        93765 Read Arounds:                         9376 (/NOCACHE)x> Read Arounds:                            0 (Function modifier)4 Read Arounds:                         9376 (IO Size)    @ 	Read aheads.  So not only does the file have to be not open forB 	writes, but VCC_MAX_IO_SIZE has to be large enough to perform theB 	read aheads (read aheads must go through cache - something else I> 	missed).  This may or may not be that big of a deal.  For me,? 	it currently isn't.  I am more constrainted on network output.t@ 	Someone writing to high bandwidth tape drives, this may be moreH 	of an issue.  There is another concern below.  I expect read aheads to B 	help more to keep the network pipe full when the pipe gets wider.  A 	If you are using Volume Shadowing - don't overlook the potentialt> 	major effect of READ_COST if NOT breaking out a volume (thereC 	is a workaround - set VCC_MAX_IO_SIZE > largest IOs :-) .  Here isi< 	backing up a 4.8 million block file, with READ_COST default 	to 1 on a 3 member shadowset:    * Total QIOs to this file:              9376* Read IOs to this file:                9376* Read Hits:                               0+ Hit Rate:                            0.00 %'* Read Aheads:                          5351* Read Throughs:                        93765 Read Arounds:                            0 (/NOCACHE)n> Read Arounds:                            0 (Function modifier)4 Read Arounds:                            0 (IO Size)  * Write IOs to this file:                  0* Write Throughs:                          05 Write Arounds:                           0 (/NOCACHE)w> Write Arounds:                           0 (Function modifier)4 Write Arounds:                           0 (IO Size)  . Average Overall I/O response time to this file,  in milliseconds:                    16.9137+ Average Disk I/O response time to this file7,  in milliseconds:                    16.91371 Accuracy of I/O resp time:                   96 %e  : 	Actual backup times and number of IOs at a ABC/TSM level:  : Direct I/O count:   87245  ! ABC/TSM is doing ~50 block IO! Data rate:          13197.27 Kb/sl! Data xfer time:     0 00:03:01.85 ! Elapsed time:       0 00:03:02.19h    : 	Questions we can answer, we know that the file is closed,< 	we know also that VCC_MAX_IO_SIZE is >= 512 blocks.  Notice= 	IO response time is 17 ms.  That's pretty good - consideringi@ 	I am pretty sure we are getting no help from the disk storage -! 	no pre-fetch help that is.  Why?   F http://www.research.compaq.com/wrl/DECarchives/DTJ/DTJ301/DTJ301SC.TXT   						 SHDRIVER attempts 						 to direct the I/O to the- 						 optimum device based on 						 locally available data.  						 This decision is based on! 						 (1) the access path, i.e.,f! 						 local or served by the VMS   						 operating system, (2) the 						 service queue lengths ate  						 the candidate controller, 						 and (3) a round-robin 						 algorithm among equal
 						 paths.n  A 	The drives in the 3 member shadowset all appear local (fibre) sot< 	skip (1).  queue lengths will stay the same (2), so we are : 	round-robining (3) and this isn't EVA so little chance ofE 	Adaptive Read Caching kicking in (Adaptive Read Caching *apparently* C 	is able to recognize patterns and I'm making an assumption that it ? 	may help if it sees groups of reads in a pattern, i.e. predict  	the next group and pre-fetch).     ? 	What about if we set VCC_MAX_IO_SIZE to the default of 127 andmB 	READ_COST is default of 1?  (Even though file is closed, setting E 	VCC_MAX_IO_SIZE to 127 will disable read-aheads as we are doing 512 lE 	block IOs out of XFC.)  We can predict that the ABC backup will run e 	slower as:g   		1)  Can't do read aheads/ 		2)  Storage can't help by pre-fetching to itsJ@ 		    cache - therefore average 512 block IO time will be higher  ; 	We are seeing that, but it is skewered by the fast writes h 	on initial creation:c  * Total QIOs to this file:             84380* Read IOs to this file:                9376* Read Hits:                               0+ Hit Rate:                            0.00 %l* Read Aheads:                             0* Read Throughs:                           05 Read Arounds:                         9376 (/NOCACHE)1> Read Arounds:                            0 (Function modifier)4 Read Arounds:                         9376 (IO Size)  * Write IOs to this file:              75004* Write Throughs:                          05 Write Arounds:                       75004 (/NOCACHE)j> Write Arounds:                       75004 (Function modifier)4 Write Arounds:                           0 (IO Size)  . Average Overall I/O response time to this file,  in milliseconds:                     3.9803+ Average Disk I/O response time to this fileO,  in milliseconds:                     3.98031 Accuracy of I/O resp time:                   81 %d  * 	However, more telling is wall clock time:  ! Buffered I/O count:            29-! Direct I/O count:           872286! Page faults:                  125m! Elapsed CPU time:   0 00:00:16.162! Data rate:          11030.97 Kb/s2! Data xfer time:     0 00:03:37.56 ! Elapsed time:       0 00:03:37.81f  A 	Repeated tests of this nature shows similar results - 20% slower ( 	(some cases 30%+ slower on other runs).  A 	Now even though we are hindered by the smaller VCC_MAX_IO_SIZE,	e: 	what if we tweek things with READ_COST?  I expect faster  	throughput, let's see.I  D $1$DGA226:     (NODE1)  ShadowSetMember      0  (member of DSAnn00:)D $1$DGA4159:    (NODE1)  ShadowSetMember      0  (member of DSAnn00:)D $1$DGA4167:    (NODE1)  ShadowSetMember      0  (member of DSAnn00:)  / NODE1.TEST> set device/read_cost=10 $1$dga4159:o/ NODE1.TEST> set device/read_cost=10 $1$dga4167:-  D 	Forcing reads to DGA226: a fibre member 1 hop away, expecting help  	from storage pre-fetch:  * Total QIOs to this file:             84380* Read IOs to this file:                9376* Read Hits:                               0+ Hit Rate:                            0.00 %o* Read Aheads:                             0* Read Throughs:                           05 Read Arounds:                         9376 (/NOCACHE)|> Read Arounds:                            0 (Function modifier)4 Read Arounds:                         9376 (IO Size)  * Write IOs to this file:              75004* Write Throughs:                          05 Write Arounds:                       75004 (/NOCACHE) > Write Arounds:                       75004 (Function modifier)4 Write Arounds:                           0 (IO Size)  . Average Overall I/O response time to this file,  in milliseconds:                     3.3716+ Average Disk I/O response time to this filec,  in milliseconds:                     3.37161 Accuracy of I/O resp time:                   80 %3  > 	What is misleading is slightly better I/O response.  But keep> 	in mind, the writes were/are skewering us.  Tale of the tape:  ! Buffered I/O count:            29i! Direct I/O count:           87224a! Page faults:                  121 ! Elapsed CPU time:   0 00:00:16.66t! Data rate:          13175.05 Kb/sT! Data xfer time:     0 00:03:02.16n! Elapsed time:       0 00:03:02.39t  C 	Faster on the wall clock - something that matters.  On much larger0> 	backups, I've seen greater disparities in timing.  Note that F 	the Elapsed time above is exactly the elapsed time if VCC_MAX_IO_SIZE? 	is set to 1000 backing up from a shadow with default READ_COSTs 	(page back up for this info):  : Direct I/O count:   87245  ! ABC/TSM is doing ~50 block IO! Data rate:          13197.27 Kb/sT! Data xfer time:     0 00:03:01.85e! Elapsed time:       0 00:03:02.19e  ? 	So practically speaking, what does this mean or what is a good 4 	example of how this will help or hurt, what to set?  > 	Since I don't want to fiddle with READ_COST on backups and I > 	do back up large files from shadowsets, files that are closed 	(log files, what not).-  > 	So I have returned VCC_MAX_IO_SIZE to 1000 and it has sped up 	several of my backups.d  @ 	If you are using older HSJ technology and are at 7.3-1 with XFC= 	turned on and you are using a 3rd party backup, you probablyu4 	want to jack VCC_MAX_IO_SIZE to see if it speeds up2 	your backups as you will not be getting pre-fetchE 	from storage controllers - even on broken out shadowsets.  (Remembert> 	if using native VMS backup, it doesn't matter Backup bypasses@ 	XFC as it does its own pre-fetching and throws nasty amounts of? 	IO at disks dependent on how tuned up your backup account is).l  = 	There is one other not so subtle advantage that XFC is doingJD 	for us, even if there isn't a an increase in performance.  Because @ 	the ABC client is doing 10 - 50 block IOs at a time (my guess, C 	reverse engineering), XFC is combining that into a 512 block IO.  eG 	This will lower interrupts on CPUs.  This is a good thing and many of tD 	us may have noticed lower kernel mode and interrupt times.  XFC is @ 	partly responsible I believe (that and re-work on IOLOCK8).  So? 	VMS backup will hammer interrupts more than third parties (32Ks 	max IO size via /BLOCK=65532).l  ; 	One final comment.  How about tweeking the crusty SHDRIVERt@ 	algorithm?  The idea there would be pick a target on sequential= 	IO.  If the IO response gets dramatically better after 3 IOs = 	(or so), then continue to send to that target.  If not, moveE9 	on to the next target, send 3 IOs, if not better move on-> 	to the 3rd target.  If after testing all three targets and noF 	improved IO, go into round robin mode.  Reasons are you may have HSJ E 	shadowed to HSG or good pre-fetching 3rd party storage.  Or you may  F 	have good third party storage or HSG80 storage and round-robining on < 	sequential IO doesn't make sense if you can force pre-fetch; 	to contoller cache.  The ugly work around for those in the9, 	know is to fiddle with READ_COST - painful.  " 	Corrections and comments, thanks!   				Robd  : And the wind shall say:  "Here were decent godless people:>                           Their only monument the asphalt road:                           And a thousand lost golf balls."-                                 -- T.S. Eliott   ------------------------------   End of INFO-VAX 2003.378 ************************