INFO-VAX Wed, 30 May 2007 Volume 2007 : Issue 293 Contents: Re: Anyone know why the Alpha market is so so quiet? RE: Anyone know why the Alpha market is so so quiet? Re: Anyone know why the Alpha market is so so quiet? Re: Anyone know why the Alpha market is so so quiet? Re: Anyone know why the Alpha market is so so quiet? Re: AUTOGEN: %SYSTEM-W-HEADERFULL, file header is full Re: AUTOGEN: %SYSTEM-W-HEADERFULL, file header is full Re: CONVERTing RMS CR/LF files to explicit byte Stream format Re: CONVERTing RMS CR/LF files to explicit byte Stream format Re: DSSI allocation class puzzle Re: DSSI allocation class puzzle Re: DSSI allocation class puzzle F$GETDVI for Disk device status MntVerifyTimeout Re: F$GETDVI for Disk device status MntVerifyTimeout Re: F$GETDVI for Disk device status MntVerifyTimeout Re: F$GETDVI for Disk device status MntVerifyTimeout Re: F$GETDVI for Disk device status MntVerifyTimeout Re: FalconStor and WAN repliaction Re: Hobbyist Licenses for ISV Layered Products Re: Is VMS losing the Financial Sector, also? new PCSI problem (8.3 Alpha) Re: OM Group acquired by Nasdaq - VMS probably out Re: openvms pascal rdb "IN" predicate parameterparsing Re: PCSI problems on 8.3 - what to do? Re: PCSI problems on 8.3 - what to do? Re: PCSI problems on 8.3 - what to do? Re: PCSI problems on 8.3 - what to do? Re: Problem with NFS client connecting to Windows 2003 server Re: RDB turorial Re: RDB Tutorial Re: RMS_TUNE_CHECK - ROOT: Primary key index root level is high Re: RMS_TUNE_CHECK - ROOT: Primary key index root level is high Re: RMS_TUNE_CHECK - ROOT: Primary key index root level is high Re: World's Fastest Processor? For Now, Anyway Re: World's Fastest Processor? For Now, Anyway Re: World's Fastest Processor? For Now, Anyway Re: World's Fastest Processor? For Now, Anyway Re: World's Fastest Processor? For Now, Anyway Re: World's Fastest Processor? For Now, Anyway Re: World's Fastest Processor? For Now, Anyway ---------------------------------------------------------------------- Date: Tue, 29 May 2007 20:04:38 +0200 From: "Dr. Dweeb" Subject: Re: Anyone know why the Alpha market is so so quiet? Message-ID: <465c6b39$0$21928$157c6196@dreader1.cybercity.dk> Main, Kerry wrote: >> -----Original Message----- >> From: William Pechter [mailto:pechter@pechter.dyndns.org] >> Sent: May 28, 2007 12:56 PM >> To: Info-VAX@Mvb.Saic.Com >> Subject: Re: Anyone know why the Alpha market is so so quiet? >> > > snip..¸ > >> >> Yup... yum update is good. >> >> What's better is a tool like I've got. >> >> Apply the patches, test. > > So is this where the full business app testing occurs or is this just > the system manager test ? > >> Make a chrooted tree with all fixes on that platform. >> Deploy with rsync to all boxes of the same type on the net. >> >> All changes from the original distribution can be rsync'd from the >> tree >> to take a kickstarted raw box (minimal system) to full functionality >> with all packages/apps installed in one upgrade rsync. >> >> Upgrade 20 machines in 10 minutes or less. >> >> This includes applying any necessary customizations for the >> applications >> where the correct configs based on machine name get sync'd with the >> updates. >> > > As I mentioned earlier, it is not the roll-out of the patches that is > the issue. Heck, that is relatively minor as you can even easily do > this with all of the Windows security patches. > > The big issue by far is the re-certification and testing of important > business applications with all of the monthly OS security patches. > > For small and some medium businesses with small numbers of users, > this is not an issue as they simply apply the patch and reboot. If a > OS security patch breaks the kernel or an application, then they can > simply roll-back with minimal impact as the numbers of users are not > that large. > > That is usually not the case with large IT environments with mission > critical environments. > > Regards > > > Kerry Main > Senior Consultant > HP Services Canada > Voice: 613-592-4660 > Fax: 613-591-4477 > kerryDOTmainAThpDOTcom > (remove the DOT's and AT) > > OpenVMS - the secure, multi-site OS that just works. OK. Just so you guys "get it", here is a real example. A system software upgrade is tested and validated. To be deployed at 8 different sites over a period of 1 year, sheduled deployment determined by PM downtime of 24*7 manufacturing operations - which by its nature is planned a long way in advance. 2 smaller sites go live before a memory leak rears its ugly head in a large site, number 3, crashing the application and stalling part of the factory shipping processes. The resulting cleanup operation consumes DBA and sysadmin time at every occurrance and occurs at different intervals depending on the transaction volume of the factory - the larger the factory, the larger the problem. We are talking daily on a large factory. The IT troubleshooters get on the job and isolate the error, create a simple reproducer and report it as priority 1 bug to the supplier, who duly fix it within 3 days! The IT guys check out the reproducer and the instances of live code where the problem was evident and verify that the supplier patch has indeed solved the problem. Q1: Which version of the software was installed at the following 5 sites? Q:2 When was the software updated at the 3 already installed sites? A1: The broken version. A2: Never (yet) In order to release a systems software upgrade, the entire application must pass certification. This is an $7B pr. year manufacturing company - a houshold name - SOX compliant and accutely aware of the necessity for application certification before deployment. Why you ask? Because the cost of bringing a larger factory down completely is like $50,000 per hour, while the cost of having a DBA cleanup the stalls is zero, because he is already sitting there and it is in his job description. The risk is evaluated, the costs apportioned and the decision made. A management no-brainer, because the certification requirement and procedures are very clear and unambiguous. As bizarre as it seems, this is the daily life of people who maintain and operate the big iron that controls large manufacturing - not just that particular site. When the application is recertified on the patched vendor software, the patch to the vendor software will be applied to the production environment in a controlled and phased manner - not before. Here endeth the lesson in reality for you guys who wouldn't know a real high-availability corporate production environment if it landed on your head! Dr. Dweeb. ------------------------------ Date: Tue, 29 May 2007 14:11:52 -0400 From: "Main, Kerry" Subject: RE: Anyone know why the Alpha market is so so quiet? Message-ID: > -----Original Message----- > From: William Pechter [mailto:pechter@pechter.dyndns.org] > Sent: May 28, 2007 12:48 PM > To: Info-VAX@Mvb.Saic.Com > Subject: Re: Anyone know why the Alpha market is so so quiet? >=20 Snip ... >=20 > I've given up this argument. Kerry's spouting the company line. > I've been there. I no longer work for a computer/os vendor. > I can speak what I see in small and medium sized IT shops. >=20 Crap. No one tells me what to say or not to say. Never has. One is expected to use good personal judgement and be professional. The same goes for any other HP person in newsgroups. I expect the same is true for most companies. > The truth is RH Linux is no different from Solaris or HP-UX from a > security > perspective and VMS doesn't run on any hardware platform at my > employer. > (We're moving off an IBM 390 series box and the rest is either AIX > (also > going away over time) and RHEL Advanced Server. >=20 For small to medium sized shops that can live with the 5-20 security patches per month, Linux (Windows) is fine. They typically do not have any requirements for re-testing applications when OS`s get patched. After each patch, the SysAdmin typically does some base level OS checks, ensures App starts up and then turns it back over to users for production. Nothing wrong with this if the business and end users are ok with the risks, then that is fine. However, large shops with mission critical apps have much higher requirements as the impact of one of these many monthly patches hosing an app or causing the system to become unstable is just to much risk to the business. Reference: (beware url wrapping) http://tinyurl.com/2jnc9y=20 http://tinyurl.com/3dxdv6 http://tinyurl.com/32tslf And this has nothing to do with OpenVMS - it is just a fact of life with large shops regardless of the platform environment. Does anyone seriously think Solaris or AIX or z-OS would be around today if they were releasing 5-20 security patches every month? Regards Kerry Main Senior Consultant HP Services Canada Voice: 613-592-4660 Fax: 613-591-4477 kerryDOTmainAThpDOTcom (remove the DOT's and AT)=20 OpenVMS - the secure, multi-site OS that just works. ------------------------------ Date: Tue, 29 May 2007 15:43:01 -0400 From: "John Smith" Subject: Re: Anyone know why the Alpha market is so so quiet? Message-ID: Main, Kerry wrote: >> -----Original Message----- >> From: William Pechter [mailto:pechter@pechter.dyndns.org] >> Sent: May 28, 2007 12:48 PM >> To: Info-VAX@Mvb.Saic.Com >> Subject: Re: Anyone know why the Alpha market is so so quiet? >> > > Snip ... > > >> >> I've given up this argument. Kerry's spouting the company line. >> I've been there. I no longer work for a computer/os vendor. >> I can speak what I see in small and medium sized IT shops. >> > > Crap. No one tells me what to say or not to say. Never has. One is > expected to use good personal judgement and be professional. The same > goes for any other HP person in newsgroups. I expect the same is true > for most companies. > >> The truth is RH Linux is no different from Solaris or HP-UX from a >> security >> perspective and VMS doesn't run on any hardware platform at my >> employer. >> (We're moving off an IBM 390 series box and the rest is either AIX >> (also >> going away over time) and RHEL Advanced Server. >> > > For small to medium sized shops that can live with the 5-20 security > patches per month, Linux (Windows) is fine. They typically do not have > any requirements for re-testing applications when OS`s get patched. > After each patch, the SysAdmin typically does some base level OS > checks, ensures App starts up and then turns it back over to users for > production. > > Nothing wrong with this if the business and end users are ok with the > risks, then that is fine. > > However, large shops with mission critical apps have much higher > requirements as the impact of one of these many monthly patches hosing > an app or causing the system to become unstable is just to much risk > to the business. > > Reference: (beware url wrapping) > > http://tinyurl.com/2jnc9y > http://tinyurl.com/3dxdv6 > http://tinyurl.com/32tslf > > And this has nothing to do with OpenVMS - it is just a fact of life > with large shops regardless of the platform environment. > > Does anyone seriously think Solaris or AIX or z-OS would be around > today if they were releasing 5-20 security patches every month? So to extend your logic, which BTW I agree with, to the lack of VMS marketing and the management levels at HP ....oh damn...there I go again... I forgot that logic regarding active promotion of VMS and HP management are mutually exclusive. whack! whack! two thumps with a 2x4 behind the woodshed for me. -- OpenVMS - The never-advertised operating system with the dwindling ISV and customer base. ------------------------------ Date: Tue, 29 May 2007 16:33:19 -0400 From: JF Mezei Subject: Re: Anyone know why the Alpha market is so so quiet? Message-ID: <80dcf$465c8e32$cef8887a$18046@TEKSAVVY.COM> Main, Kerry wrote: > For small to medium sized shops that can live with the 5-20 security > patches per month, Linux (Windows) is fine. If you install the latest and greatest, there may be a number of monthly bugs that are discovered and patches issued for them. But if you istall an older version, a lot of it will have been debugged and you take the original media, apply a whole set of patches that were issued while that product was "new" and you end up with a stable OS. In other words, Linux is released as beta, during which time, you have a whole buch of patches needed. Eventually, that version stabilises and then becomes ready to move from hobbyist and early adopter to more serious uses. ------------------------------ Date: Tue, 29 May 2007 22:29:39 -0400 From: =?ISO-8859-1?Q?Arne_Vajh=F8j?= Subject: Re: Anyone know why the Alpha market is so so quiet? Message-ID: <465ce193$0$90271$14726298@news.sunsite.dk> Ron Johnson wrote: > On 05/28/07 15:58, Arne Vajhøj wrote: >> Main, Kerry wrote: >>> Software requires time and skilled resources. Both are limited >>> resources. As the old saying goes - "Linux is free as long as you >>> value your time as free as well." >> >> But those subscribing to that philosophy will in 20 years >> either be working on IBM mainframe or flipping burgers. >> >> It is the equivalent of 12-14 years ago when I heard a >> DEC reseller when asked about whether it was smart of DEC >> to make an alliance with MS replied "MS is just a small company - who >> do they think they are.". > > Except now DEC is gone and it's products are gone or disappearing, and > MSFT is the most profitable company in the known multiverse. That is not "except" - it is the main point !!!! Arne ------------------------------ Date: Tue, 29 May 2007 18:45:29 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: AUTOGEN: %SYSTEM-W-HEADERFULL, file header is full Message-ID: JF Mezei writes: >$SET DIRECTORY SYS$SPECIFIC:[TCPIP$SMTP]/VERSION=10 >$PURGE SYS$SPECIFIC:[TCPIP$SMTP]TCPIP$SMTP_RECV_RUN.LOG >$set file SYS$SPECIFIC:[TCPIP$SMTP]TCPIP$SMTP_RECV_RUN.LOG;*/version=10 Due to the way ODS-2 works, the last statement doesn't need the ";*", there is only version limit field for all of the versions of a given file. You do need to do a purge to get down to the limit (second statement) since only one version gets deleted when a new version gets created, no matter how many would need to be deleted to get down to the version limit. ------------------------------ Date: Tue, 29 May 2007 19:12:16 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: AUTOGEN: %SYSTEM-W-HEADERFULL, file header is full Message-ID: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply) writes: >When running AUTOGEN, I got the dreaded >%SYSTEM-W-HEADERFULL, file header is full > ** WARNING ** - Error creating SYS$SYSROOT:[SYSEXE]SYSDUMP.DMP. > ** ERROR ** - error creating SYSDUMP.DMP. SYSDUMP.DMP needs > to be created manually with 393258 blocks >Presumably, I would get the same error when trying to create it >manually, right? >$ dir SYS$SYSROOT:[SYSEXE]SYSDUMP.DMP/siz >Directory SYS$SYSROOT:[SYSEXE] >SYSDUMP.DMP;4 215910 >OK, it wants a larger dump file, and there is room: ... Many people have mentioned that INDEXF.SYS has filled its single header. It is true that INDEXF.SYS is limited to a single header and attempts to create more files that would cause it to exceed this will result in the HEADERFULL error. SYSDUMP.DMP (and %%%%FILE.SYS) also have this single header limitation, and attempts to expand these files by lots of little chunks will also get the HEADERFULL failure. Your disk is fragmented. You need a few largish chunks of contiguous space to expand SYSDUMP.DMP, not lots and lots of tiny chunks. An image backup/restore of the system disk will fix this. A VMS disk defragger will as well. Trial-and-error way is to declare war on unneeded files, getting rid of old log files, restarting things like audit server log files & deleting old ones (you have 4,000+ daily process creations from spam alone, and likely things like SSH breakin attempts to fill it up), then trying again. Your sysdump file header may be full from previous expansions, and you can't expand it by even one block now. Try creating a new dump file: SYSGEN> CREATE SYS$SYSROOT:[SYSEXE]SYSDUMP.DMP /SIZE=nnnnn as a new version, and delete the old version after the next reboot. ------------------------------ Date: Wed, 30 May 2007 08:04:59 +0800 From: "Richard Maher" Subject: Re: CONVERTing RMS CR/LF files to explicit byte Stream format Message-ID: Hi Jeff, I thought I'd already tried that and at first couldn't get your example to work here until I realized that I was attempting to convert an earlier pre-altered test file that contained delimiters in the data :-( Works great now! thanks a lot. Cheers Richard Maher "Jeffrey H. Coffield" wrote in message news:NSV6i.8488$4Y.7226@newssvr19.news.prodigy.net... > Richard Maher wrote: > > Hi, > > > > This has got to be easy but after floundering around for an hour I can't > > find it, so could someone please tell me how to take a variable length > > carriage control file and remove any implicit RMS processing and control > > info, so that the end result is a "What you see is What you get" UDF file of > > bytes? > > > > In other words if I have a Var Length CR control file, with two records > > > > hello > > world > > > > Then I'd like to end up with a Record Format = Undefined and Record > > Attributes=None > > > > helloworld > > > > I've played around with convert and FDL and can't get it to work. I don't > > have to write a program for this do I? > > > > Why? I want to perform block i/o on the file and will lose any rms record > > attributes. > > > > Cheers Richard Maher > > > > > > > > $ dump hw.txt > > Virtual block number 1 (00000001), 512 (0200) bytes > > 00646C72 6F770005 006F6C6C 65680005 ..hello...world. 000000 > 00000000 00000000 00000000 0000FFFF ................ 000010 > ... > > $ type hw.fdl > > IDENT FDL_VERSION 02 "29-MAY-2007 06:20:24 OpenVMS FDL Editor" > > FILE > ORGANIZATION sequential > > RECORD > BLOCK_SPAN yes > CARRIAGE_CONTROL none > FORMAT stream > > $ convert/fdl=hw.fdl hw.seq > > $ dump hw.seq > > Virtual block number 1 (00000001), 512 (0200) bytes > > 00000A0D 646C726F 770A0D6F 6C6C6568 hello..world.... 000000 > 00000000 00000000 00000000 00000000 ................ 000010 > > Jeff Coffield ------------------------------ Date: Wed, 30 May 2007 08:21:05 +0800 From: "Richard Maher" Subject: Re: CONVERTing RMS CR/LF files to explicit byte Stream format Message-ID: Hi David, > Alternately: > > $ COPY NLA0: STREAM_FILE.DAT > $ SET FILE/ATTR=RFM=STM > $ APPEND MYFILE.TXT STREAM_FILE.DAT > $ SET FILE/ATTR=RFM=UDF > > ...should do what you want. That seems to work too, even with the warning message, thanks. I'm gonna skip the UDF bit for the moment 'cos I was really more interested in being able to TSTB FAB$B_RAT and return an error if non-zeros but it looks like CARRIAGE_CONTROL just gets stuck there for stream files. I know I can override this with set file but it's all a bit messy. So rather than try to vet everything that's going on, I'll just dump what's physically on the disk. If someone wants implicit enrichment or value added processing then tough-titties (at least for the moment). Cheers Richard Maher "David J Dachtera" wrote in message news:465C5F62.CE6819B4@spam.comcast.net... > "Jeffrey H. Coffield" wrote: > > > > Richard Maher wrote: > > > Hi, > > > > > > This has got to be easy but after floundering around for an hour I can't > > > find it, so could someone please tell me how to take a variable length > > > carriage control file and remove any implicit RMS processing and control > > > info, so that the end result is a "What you see is What you get" UDF file of > > > bytes? > > > > > > In other words if I have a Var Length CR control file, with two records > > > > > > hello > > > world > > > > > > Then I'd like to end up with a Record Format = Undefined and Record > > > Attributes=None > > > > > > helloworld > > > > > > I've played around with convert and FDL and can't get it to work. I don't > > > have to write a program for this do I? > > > > > > Why? I want to perform block i/o on the file and will lose any rms record > > > attributes. > > > > > > Cheers Richard Maher > > > > > > > > > > > > > $ dump hw.txt > > > > Virtual block number 1 (00000001), 512 (0200) bytes > > > > 00646C72 6F770005 006F6C6C 65680005 ..hello...world. 000000 > > 00000000 00000000 00000000 0000FFFF ................ 000010 > > ... > > > > $ type hw.fdl > > > > IDENT FDL_VERSION 02 "29-MAY-2007 06:20:24 OpenVMS FDL Editor" > > > > FILE > > ORGANIZATION sequential > > > > RECORD > > BLOCK_SPAN yes > > CARRIAGE_CONTROL none > > FORMAT stream > > > > $ convert/fdl=hw.fdl hw.seq > > > > $ dump hw.seq > > > > Virtual block number 1 (00000001), 512 (0200) bytes > > > > 00000A0D 646C726F 770A0D6F 6C6C6568 hello..world.... 000000 > > 00000000 00000000 00000000 00000000 ................ 000010 > > > > Jeff Coffield > > Alternately: > > $ COPY NLA0: STREAM_FILE.DAT > $ SET FILE/ATTR=RFM=STM > $ APPEND MYFILE.TXT STREAM_FILE.DAT > $ SET FILE/ATTR=RFM=UDF > > ...should do what you want. > > -- > David J Dachtera > dba DJE Systems > http://www.djesys.com/ > > Unofficial OpenVMS Marketing Home Page > http://www.djesys.com/vms/market/ > > Unofficial Affordable OpenVMS Home Page: > http://www.djesys.com/vms/soho/ > > Unofficial OpenVMS-IA32 Home Page: > http://www.djesys.com/vms/ia32/ > > Unofficial OpenVMS Hobbyist Support Page: > http://www.djesys.com/vms/support/ ------------------------------ Date: Tue, 29 May 2007 15:04:03 -0400 From: JF Mezei Subject: Re: DSSI allocation class puzzle Message-ID: gerry77@no.spam.mail.com wrote: > Actually I've got only _one_ DSSI cable (I'm just a hobbyist) so I can > connect at most one bus between hosts. Which bus do you suggest? The one > with the disks (bus 0) or the other one? There would be any performance > difference? Connect the bus with the disks. That is a HUGE difference. This way, both nodes have direct access to the drives at the DSSI level, even if the other node is down (but powered up). If you have the cable on bus 1, and each node has disks on bus0, then when VAXA needs to access a drive on VAXB's bus 0, it needs to use Bus1 to send and MSCP message to the MSCP process on VAXB which then accesses the data and sends it back via BUS1 to VAXA. If VAXB is down, you cannot access that drive from VAXA. (and this scheme also adds to the CPU load on both nodes since they need to process each other's disk access requests. ------------------------------ Date: Tue, 29 May 2007 15:30:00 -0400 From: JF Mezei Subject: Re: DSSI allocation class puzzle Message-ID: <8f305$465c7f5c$cef8887a$1338@TEKSAVVY.COM> gerry77@no.spam.mail.com wrote: > However I'm wondering if this is still true when the UNITNUM is set to > any number higher than 7. The "electrical" ID set with plugs should be > quite low level if compared with MSCP and friends: are you saying that > during its initialization the controller does a bus scan for every > possibile ID, even those higher than 7? Is that a too SCSI-oriented > thought? http://www.mcmanis.com/chuck/computers/vaxen/dssi-plug.html UNITNUM does not change the "SCSI ID" of the drive at the bus level. The IFPlug (or dip switches in the back) do that. UNITNUM changes how the drive "name" looks like, not how the drive is adressed on the bus. ------------------------------ Date: Tue, 29 May 2007 20:39:30 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: DSSI allocation class puzzle Message-ID: JF Mezei writes: >gerry77@no.spam.mail.com wrote: >> Actually I've got only _one_ DSSI cable (I'm just a hobbyist) so I can >> connect at most one bus between hosts. Which bus do you suggest? The one >> with the disks (bus 0) or the other one? There would be any performance >> difference? >Connect the bus with the disks. That is a HUGE difference. >This way, both nodes have direct access to the drives at the DSSI level, >even if the other node is down (but powered up). >If you have the cable on bus 1, and each node has disks on bus0, then >when VAXA needs to access a drive on VAXB's bus 0, it needs to use Bus1 >to send and MSCP message to the MSCP process on VAXB which then accesses >the data and sends it back via BUS1 to VAXA. If VAXB is down, you cannot >access that drive from VAXA. (and this scheme also adds to the CPU load >on both nodes since they need to process each other's disk access requests. In addition, newer version of VMS will spread SCS traffic over all paths available to it, and will favor the path with the least latency (fastest). More paths = more possibilities. However, DSSI is slow, not much to do about that. Also, I'm not 100% sure the speedup code is on the VAX SCS code, with the most recent being V7.3. But still, direct access on Bus X will be faster than Bus X -> VAXB -> Bus Y via MSCP Serving. ------------------------------ Date: Tue, 29 May 2007 14:34:12 -0400 From: norm.raphael@metso.com Subject: F$GETDVI for Disk device status MntVerifyTimeout Message-ID: Is there a call to F$GETDVI for Disk device status MntVerifyTimeout? ------------------------------ Date: 29 May 2007 15:31:53 -0500 From: brooks@cuebid.zko.hp.nospam (Rob Brooks) Subject: Re: F$GETDVI for Disk device status MntVerifyTimeout Message-ID: norm.raphael@metso.com writes: > Is there a call to F$GETDVI for Disk device status MntVerifyTimeout? No. There is no explicit status that reflects that state. A device is said to have timed out of mount verification if all the below conditions are true: 1) The device is mounted (bit ucb$v_mnt) 2) The device is not in mount verification (bit ucb$v_mntverip) 3) The valid bit is clear (bit ucb$v_valid) ucb$v_mnt is a bit in the first device characteristics longword (ucb$l_devchar) ucb$v_mntverip and ucb$v_valid are bits in the device status longword (ucb$l_sts) SDA> SHOW DEVICE is a good way to look at the gory details -- Rob Brooks MSL -- Nashua brooks!cuebid.zko.hp.com ------------------------------ Date: Tue, 29 May 2007 18:30:02 -0400 From: norm.raphael@metso.com Subject: Re: F$GETDVI for Disk device status MntVerifyTimeout Message-ID: brooks@cuebid.zko.hp.nospam (Rob Brooks) wrote on 05/29/2007 04:31:53 PM: > norm.raphael@metso.com writes: > > > Is there a call to F$GETDVI for Disk device status MntVerifyTimeout? > > No. There is no explicit status that reflects that state. > > A device is said to have timed out of mount verification if all the > below conditions are true: > > 1) The device is mounted (bit ucb$v_mnt) Good > 2) The device is not in mount verification (bit ucb$v_mntverip) Good > 3) The valid bit is clear (bit ucb$v_valid) Good > This seems to me to be a mounted, valid disk (a good one), not one in MntVerifyTimeout. What am I missing? This one is not timeout: Device status: 88021810 online,VALID,unload,lcl_valid,exfunc_supp,iopost_local Characteristics: 1C4D4008 dir,fod,shr,avl,MNT,elg,idv,odv,rnd 03082021 clu,mscp,loc,vrt,scsi,wlg > ucb$v_mnt is a bit in the first device characteristics longword > (ucb$l_devchar) > > ucb$v_mntverip and ucb$v_valid are bits in the device status longword > (ucb$l_sts) > > SDA> SHOW DEVICE is a good way to look at the gory details > > -- > > Rob Brooks MSL -- Nashua brooks!cuebid.zko.hp.com ------------------------------ Date: Wed, 30 May 2007 01:46:13 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: F$GETDVI for Disk device status MntVerifyTimeout Message-ID: norm.raphael@metso.com writes: >brooks@cuebid.zko.hp.nospam (Rob Brooks) wrote on 05/29/2007 04:31:53 PM: >> A device is said to have timed out of mount verification if all the >> below conditions are true: ... >> 3) The valid bit is clear (bit ucb$v_valid) >This seems to me to be a mounted, valid disk (a good one), >not one in MntVerifyTimeout. What am I missing? >This one is not timeout: >Device status: >88021810 online,VALID,unload,lcl_valid,exfunc_supp,iopost_local >Characteristics: >1C4D4008 dir,fod,shr,avl,MNT,elg,idv,odv,rnd >03082021 clu,mscp,loc,vrt,scsi,wlg He said Valid _clear_. You show one with Valid _set_, which is a usable mounted disk. ------------------------------ Date: Tue, 29 May 2007 21:59:09 -0400 From: norm.raphael@metso.com Subject: Re: F$GETDVI for Disk device status MntVerifyTimeout Message-ID: This is a multipart message in MIME format. --=_alternative 000AE89E852572EB_= Content-Type: text/plain; charset="US-ASCII" moroney@world.std.spaamtrap.com (Michael Moroney) wrote on 05/29/2007 09:46:13 PM: > norm.raphael@metso.com writes: > > >brooks@cuebid.zko.hp.nospam (Rob Brooks) wrote on 05/29/2007 04:31:53 PM: > > >> A device is said to have timed out of mount verification if all the > >> below conditions are true: > ... > > >> 3) The valid bit is clear (bit ucb$v_valid) > > >This seems to me to be a mounted, valid disk (a good one), > >not one in MntVerifyTimeout. What am I missing? > > >This one is not timeout: > >Device status: > >88021810 online,VALID,unload,lcl_valid,exfunc_supp,iopost_local > >Characteristics: > >1C4D4008 dir,fod,shr,avl,MNT,elg,idv,odv,rnd > >03082021 clu,mscp,loc,vrt,scsi,wlg > > He said Valid _clear_. You show one with Valid _set_, which is a usable > mounted disk. Thank you. [I plead a senior moment. I really _can_ read.] Is there a DCL trick to access these longwords? --=_alternative 000AE89E852572EB_= Content-Type: text/html; charset="US-ASCII"



moroney@world.std.spaamtrap.com (Michael Moroney) wrote on 05/29/2007 09:46:13 PM:

> norm.raphael@metso.com writes:
>
> >brooks@cuebid.zko.hp.nospam (Rob Brooks) wrote on 05/29/2007 04:31:53 PM:
>
> >> A device is said to have timed out of mount verification if all the
> >> below conditions are true:
> ...
>
> >> 3) The valid bit is clear (bit ucb$v_valid)
>
> >This seems to me to be a mounted, valid disk (a good one),
> >not one in MntVerifyTimeout.  What am I missing?
>
> >This one is not timeout:
> >Device status:
> >88021810 online,VALID,unload,lcl_valid,exfunc_supp,iopost_local
> >Characteristics:
> >1C4D4008 dir,fod,shr,avl,MNT,elg,idv,odv,rnd
> >03082021 clu,mscp,loc,vrt,scsi,wlg
>
> He said Valid _clear_.  You show one with Valid _set_, which is a usable
> mounted disk.


Thank you.  [I plead a senior moment.  I really _can_ read.]

Is there a DCL trick to access these longwords?
--=_alternative 000AE89E852572EB_=-- ------------------------------ Date: 29 May 2007 11:03:20 -0700 From: BaxterD@tessco.com Subject: Re: FalconStor and WAN repliaction Message-ID: <1180461800.243889.158140@g4g2000hsf.googlegroups.com> On May 29, 12:59 pm, David J Dachtera wrote: > Gorazd Kikelj wrote: > > > Hi TOm, > > > will need to ask customer. It was their setup that didn't involve as :-( > > > In core you need to present volumes as HSG80 volumes and it should work. > > Customer will send me a procedure in a near future. > > > Also I see on FalconStore site, that OpenVMS is listed as supported OS > > platform. > > Do you, perchance, have a URL where I could find that? The folks at work would > be interested... (please: a -SPECIFIC- URL, not justhttp://www.falconstor.com/ > ) > > -- > David J Dachtera > dba DJE Systemshttp://www.djesys.com/ > > Unofficial OpenVMS Marketing Home Pagehttp://www.djesys.com/vms/market/ > > Unofficial Affordable OpenVMS Home Page:http://www.djesys.com/vms/soho/ > > Unofficial OpenVMS-IA32 Home Page:http://www.djesys.com/vms/ia32/ > > Unofficial OpenVMS Hobbyist Support Page:http://www.djesys.com/vms/support/ Hi Dave, Although I got this off a "SUN" Site, it is a FalconStor Doc. Follow the URL and check out the client table at the bottom of the PDF. http://www.sun.com/storagetek/management_software/data_protection/ipstor/datasheet.pdf Dave Baxter. ------------------------------ Date: 29 May 2007 14:35:42 -0500 From: Kilgallen@SpamCop.net (Larry Kilgallen) Subject: Re: Hobbyist Licenses for ISV Layered Products Message-ID: <56VkO6V0Wnt6@eisner.encompasserve.org> In article <465A4FD1.2F25054E@spam.comcast.net>, David J Dachtera writes: > Richard Maher wrote: >> We're in the process of coming up with a Hobbyist License agreement for >> Tier3/hotTIP and was wondering if someone who's already created a Hobbyist >> License for a third-party product could answer a couple of questions: - >> >> - It appears that it is usual to piggy-back the agreement on the >> Decus/Encompass/HP-VMS hobbyist license, is that correct? Is having a VMS >> Hobbyist agreement normally a pre-condition for issuing a hobbyist license >> for a third-party product? Being an individual producer does not give one special credentials for answering question about overall trends. Since Hobbyist Licenses are not a revenue source, there is no incentive to study how other producers do it. I think you will need to conduct your own survey. >> - If so then how is compliance achieved? Using LMF to check for a single >> OPENVMS-HOBBYIST PAK? Yes (in our case). >> - Where can I get a copy of the exact wording of your Hobbyist License >> Agreement? (I have a copy of the old "COMPAQ HOBBY LICENSE AGREEMENT For >> OpenVMS" the T's & C's of which appear to be mirrored in other ISV >> agreements. Is this the norm?) http://www.ljk.com/ljk/LJK_CDROM_DOCUMENTATION/ljk_cdrom_0100_006.html#licenses_hobbyist I found that with a Google search for the word Hobbyist on our site. ------------------------------ Date: Tue, 29 May 2007 15:34:59 -0400 From: "John Smith" Subject: Re: Is VMS losing the Financial Sector, also? Message-ID: <957b6$465c805b$cef89dd0$2949@TEKSAVVY.COM-Free> Ron Johnson wrote: > On 05/28/07 10:21, John Smith wrote: >> Not surprising. >> >> If you need any help on the conversion, we have some people on-staff >> who used to be with Forte's software engineering & consulting team >> and who have a bunch of Siebel/Oracle experience. >> >> Some of my customers were amongst the first non-beta V1.0 sites >> Forte had. Always thought it was a great product - incredibly >> productive to code in and flexible for the server side - unlike that >> 3gl write once-crash anywhere Java garbage. > > I'm sure that the app wasn't implemented well (no one in our group > had any OO experience back in 1998), and the framework that > consultants set up was a disaster. > > Contact me off-line and I'll try to put you in touch with the right > person. One of my partners will contact you either later today or tomorrow. -- OpenVMS - The never-advertised operating system with the dwindling ISV and customer base. ------------------------------ Date: Wed, 30 May 2007 00:51:52 GMT From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) Subject: new PCSI problem (8.3 Alpha) Message-ID: <00A685A2.8104BE0D@SSRL.SLAC.STANFORD.EDU> DS20E VMS 8.3 Alpha -- ----------------------------------------------------------------------- The following product will be installed to destination: DEC AXPVMS VMS83A_UPDATE V2.0 DISK$AXPRASYS:[VMS$COMMON.] %PCSI-F-ASSERTFAIL, assertion failure: Exe par internal error at line 2536 in $1$DGA1037:[PCSI_3.SRC]SPIU_ORDER_EXECUTION.C;1 %PCSI-E-OPFAILED, operation failed ----------------------------------------------------------------------- ITRC showed a dependency on an 82 PCSI update, but that (obviously) couldn't be installed. -- Alan ------------------------------ Date: Tue, 29 May 2007 16:25:10 -0400 From: JF Mezei Subject: Re: OM Group acquired by Nasdaq - VMS probably out Message-ID: John Smith wrote: > Frankly I'd be amazed if there was *anything* exhaustive analyzed knowing > how M&A deals are often done. "True" due diligence cannot be done until the deal has been announced publically. Before that, you get Carly and Curly discussing their wedding in front of a fireplace with a champagne and exchanging what they know and VERY FEW other people within the respective corporations would be in the loop to provide them with information on the impact of such a takeover/merger. M&A bankers will advise Carly/Curly on possible synergies of such a merger based only on publically available documents from both companies. They can make broad statements about cost savings etc, but they cannot make definitive ones. Once the deal is publically announced, Carly and Curly drop their clothes and have access to each other's books, and they can they designate key staff within each company to do the same and report back to them. Such key staff, by having intimate access to a competitor's books/operations/secrets risks losing his/her job should the merger fail because of the competitive information they have acquired. But this only happens once the deal is announced. So in the case of NASDAQ, their statement that apparently claims that their technology is superior to OM's is just the NASDAQ chairmain making an uninformed opinion to help make his dick appear bigger. Now that the deal is announced, due diligence can happen and they can then make more informed guesses on what Nasdaq would actually be buying and whether the OM platform is worth preserving or not. And this is why its i extremely important for OM and the VMS management within HP to ensure NASDAQ is given high level "VMS is the best" presentations. ------------------------------ Date: Tue, 29 May 2007 20:08:29 +0200 From: "Dr. Dweeb" Subject: Re: openvms pascal rdb "IN" predicate parameterparsing Message-ID: <465c6c20$0$21928$157c6196@dreader1.cybercity.dk> Ron Johnson wrote: > On 05/29/07 04:45, wim.de-boer@corusgroup.com wrote: >> Hello PASCAL / RDB-friends, >> Working on openVMS 8.2 on Alpha DS15 using Pascal V5.9-95 and Oracle >> Rdb V7.1-451. >> >> I'm trying to pass a varying amount of strings to an IN-predicate, >> please see the example > > Ah, I don't think you can do that. > > I would suggest using a global temporary table. > That's what I'd do, but then I am a bit lazy in these matters. Dr. Dweeb > At the SQL> prompt, > CREATE GLOBAL TEMPORARY TABLE GTT_FOOBAR ( > VAL CHAR(10)) > ON COMMIT PRESERVE ROWS; > > Then, in your program: > DECLARE Bezette_vakken_in_rij READ ONLY CURSOR FOR > SELECT > matr_park_lok_x, > matr_park_lok_y > FROM materiaal_eenheid WHERE > matr_park_id = :Park AND > matr_proces_sts = 'IN' AND > matr_park_lok_x IN (SELECT SOME_VAL FROM GTT_FOOBAR) > GROUP BY matr_park_lok_x, matr_park_lok_y > > Before you call Open_bezette_vakken_in_rij, delete from GTT_FOOBAR, > then make a loop that inserts all the necessary records into > GTT_FOOBAR; ------------------------------ Date: Tue, 29 May 2007 22:04:55 GMT From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) Subject: Re: PCSI problems on 8.3 - what to do? Message-ID: <00A6858B.2E45CDFC@SSRL.SLAC.STANFORD.EDU> In article <1180429978.830693.169410@p77g2000hsh.googlegroups.com>, IanMiller writes: >what version of unzip ? UnZip 5.20 of 30 April 1996, by Info-ZIP. Maintained by Greg Roelofs. Send bug reports to the authors at Zip-Bugs@wkuvx1.wku.edu; see README for details. Latest sources and executables are always in ftp.uu.net:/pub/archiving/zip, at least as of date of this release; see "Where" for other ftp and non-ftp sites. Compiled with DEC C V5.2-003 for OpenVMS (V6.2 for Alpha) on May 1 1996. UnZip special compilation options: COPYRIGHT_CLEAN (PKZIP 0.9x unreducing method not supported) [decryption, version 2.6 of 23 April 1996] UnZip and ZipInfo environment options: UNZIP_OPTS: [none] UNZIPOPT: [none] ZIPINFO_OPTS: [none] ZIPINFOOPT: [none] -- Alan ------------------------------ Date: Tue, 29 May 2007 17:19:24 -0500 (CDT) From: sms@antinode.org (Steven M. Schweda) Subject: Re: PCSI problems on 8.3 - what to do? Message-ID: <07052917192463_202002DA@antinode.org> From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) > In article <1180429978.830693.169410@p77g2000hsh.googlegroups.com>, IanMiller writes: > >what version of unzip ? > > UnZip 5.20 of 30 April 1996, by Info-ZIP. [...] Yikes. What have you been doing for the past decade? Try something less archaic, like, say: UnZip 5.52 of 28 February 2005, by Info-ZIP. [...] http://www.info-zip.org/ http://www.info-zip.org/UnZip.html http://www.info-zip.org/Zip.html Probably be faster, too. ------------------------------------------------------------------------ Steven M. Schweda sms@antinode-org 382 South Warwick Street (+1) 651-699-9818 Saint Paul MN 55105-2547 ------------------------------ Date: Wed, 30 May 2007 00:25:14 GMT From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) Subject: Re: PCSI problems on 8.3 - what to do? Message-ID: <00A6859E.C86715D3@SSRL.SLAC.STANFORD.EDU> In article <07052917192463_202002DA@antinode.org>, sms@antinode.org (Steven M. Schweda) writes: >From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) > >> In article <1180429978.830693.169410@p77g2000hsh.googlegroups.com>, IanMiller writes: >> >what version of unzip ? >> >> UnZip 5.20 of 30 April 1996, by Info-ZIP. [...] > > Yikes. What have you been doing for the past decade? Try something >less archaic, like, say: > >UnZip 5.52 of 28 February 2005, by Info-ZIP. [...] > > > http://www.info-zip.org/ > http://www.info-zip.org/UnZip.html > http://www.info-zip.org/Zip.html > > Probably be faster, too. Embarrassingly enough, I had 5.42 (sic) installed but had an archaic symbol definition pointing me to 5.20. 5.42 (on Itanium) had the same complaints. Have now installed 5.52 and zapped the old symbol definition. Ran it again, got the same problem, which wasn't related to ZIP version. There's something hosed with saving the ZIP file to VMS disk over Samba: openvms_alpha_8_3_05290535.zip;2 49392/49392 29-MAY-2007 17:03:53.58 (RWED,RWED,R,) openvms_alpha_8_3_05290535.zip;1 29536/29538 29-MAY-2007 16:45:43.34 (RWED,RWED,R,) (version 1 was with Samba; version 2 was saved to PC desktop and FTPed over, and unzip throws no errors on that.) -- Alan ------------------------------ Date: Tue, 29 May 2007 21:08:28 -0500 (CDT) From: sms@antinode.org (Steven M. Schweda) Subject: Re: PCSI problems on 8.3 - what to do? Message-ID: <07052921082891_202002DA@antinode.org> From: winston@SSRL.SLAC.STANFORD.EDU (Alan Winston - SSRL Central Computing) > >> UnZip 5.20 of 30 April 1996, by Info-ZIP. [...] > >UnZip 5.52 of 28 February 2005, by Info-ZIP. [...] > Embarrassingly enough, I had 5.42 (sic) installed but had an archaic symbol > definition pointing me to 5.20. 5.42 (on Itanium) had the same complaints. > Have now installed 5.52 and zapped the old symbol definition. I feel better now. (Zip 3 and UnZip 6 are still threatening to become available before we all die, too, so stay alert for the next great leap forward.) > Ran it again, got the same problem, which wasn't related to ZIP version. > There's something hosed with saving the ZIP file to VMS disk over Samba: > > openvms_alpha_8_3_05290535.zip;2 > 49392/49392 29-MAY-2007 17:03:53.58 (RWED,RWED,R,) > openvms_alpha_8_3_05290535.zip;1 > 29536/29538 29-MAY-2007 16:45:43.34 (RWED,RWED,R,) > > (version 1 was with Samba; version 2 was saved to PC desktop and FTPed over, > and unzip throws no errors on that.) Ok. I can understand why you would prefer not to disclose the UnZip version, but it's less clear why you wouldn't want to show the DIRE /FULL results on the archives. Also, "My unZIP thought it was corrupt" is not a very useful description of the actual error message from UnZip. On the other hand, with that kind of size difference, it's clear that something Samba-related is majorly hosed, so potentially useful info may not actually tell us much. Luckily, with the low density of Windows systems here, I don't need to worry much about Samba, so my expertise there is nearly nil. Everything's complicated. Sigh. ------------------------------------------------------------------------ Steven M. Schweda sms@antinode-org 382 South Warwick Street (+1) 651-699-9818 Saint Paul MN 55105-2547 ------------------------------ Date: Wed, 30 May 2007 08:18:24 +1000 From: "Gremlin" Subject: Re: Problem with NFS client connecting to Windows 2003 server Message-ID: <135p9ljbr5h1e42@corp.supernews.com> Hi All Well, it was almost that...... Thanks to you both, the information you gave me didn't work directly, but it made me think (always a hard thing!). There is a well known (not to me) bug that says that the NFS authentication server must *also* run on a Domain Controller or the whole thing doesn't work. So, I installed the authentication server on the DC as well as the other windows server and - voila!! Thanks for all your help "Gorazd Kikelj" wrote in message news:f3h9mc$g0u$1@usenet01.boi.hp.com... > > "Gremlin" wrote in message > news:135nf8i6dheh60a@corp.supernews.com... >> Hi All >> >> [SYSMGR] > tcpip sho ver >> >> HP TCP/IP Services for OpenVMS Alpha Version V5.5 - ECO 1 >> on an AlphaServer DS10L 466 MHz running OpenVMS V8.2 >> >> I have created an NFS share on a Windows2003 server running Windows >> Services for Unix v3.5. The share is OK, there are no error messages in >> the Windows log file, I have mounted the share using: >> >> TCPIP> mount dnsf0: /host="DL380" /path="NFSdata" >> %TCPIP$DNFSMOUNT-S-MOUNTED, NFSdata mounted on _DNFS7:[000000] >> > ... >> [SYSMGR] > dir dnfs7:[000000] >> %DIRECT-E-OPENIN, error opening DNFS7:[000000]*.*;* as input >> -RMS-E-DNF, directory not found >> -SYSTEM-F-TIMEOUT, device timeout >> >> Can't see it, get to it, etc. >> >> Any thoughts appreciated - cheers >> > > This is usualy missing username mappings and proxy settings, hence > security. > So create a valid username mapping on SFU (Windows) side and then also > create right proxy records to reflect win mappings. And you are done :-) > > > So create a mapping in SFU for example: > > uid=100, gid=100 -> windows\Administrator > > and proxy in tcpip > > uid=100, gid=100, vmsuser > > that should do the trick. > > Best, Gorazd > ------------------------------ Date: Tue, 29 May 2007 20:07:08 +0200 From: "Dr. Dweeb" Subject: Re: RDB turorial Message-ID: <465c6bce$0$21932$157c6196@dreader1.cybercity.dk> Ruslan R. Laishev wrote: > Hello, Hal! > > > Hal Kuff wrote: >> There are quite a few (three) GREAT books from the digital press... >> read Them!!!! In fact Lilian Hobbs, the creator or one of the >> creators of rdb wrote them. > Thanks! I'll try to find it. > If you have a problem I will do a search for you, but I think you will be able to find the reference. I am not sure of the current status of the Hobbes et. al. book or whether it is still in print. Dr. Dweeb (a hardened veteran Rdb'er) >> Also, how the hell are you these days? Still doing kerberos? > Kerberos ... RADIUS ... teaching young VMS students at Ford Russia, > and so on... > >> >> >> "Ruslan R. Laishev" >> wrote in >> message news:3995DF2D429C1314ED2D60760705A2C1@NNTP.DeltaTel.RU... >>> Hello, All! >>> >>> >>> I looking for RDB tutorial "for beginners". >>> >>> >>> Thanks for any pointers. >>> >>> >>> -- >>> + WBR, OpenVMS [Sys|Net] HardWorker ............. Skype: SysMan-One >>> + Delta Telecom JSC, IMT-MC-450(CDMA2000) cellular operator >>> Russia,191119,St.Petersburg,Transportny per. 3 Cel: +7 (812) >>> 716-3222 +http://starlet.deltatelecom.ru ............. Frying on >>> OpenVMS only + ------------------------------ Date: Wed, 30 May 2007 09:09:13 +0500 From: Valentin Likoum Subject: Re: RDB Tutorial Message-ID: <1987160271.20070530090913@ncc.volga.ru> On 29/05/07 Ruslan R. Laishev wrote: > Hello, Dr. Dweeb! > Dr. Dweeb wrote: >> Ruslan R. Laishev wrote: >>> Hello, All! >>> >>> >>> I looking for RDB tutorial "for beginners". >>> >>> >>> Thanks for any pointers. >> >> Do you mean OracleRdb as a specific product, or "rdb" as a generic reference >> to relational databases. >> >> Read Chris Date's book for the latter and I guess the Hobbs book for the >> former. >> Amazon search is your friend here. > I looking for any guide which describe for beginners several ways for > developing/programming apps under/on RDB: > - SQL Mods > - SQL Preprocs > - Stored Procedures, Packages (?) > - Creating and using constants in the stored > procedures/packages (as is it in > the ORACLE PL/SQL stuff). Obviously "Oracle Rdb7 Guide to SQL Programming" http://download.oracle.com/otn_hosted_doc/rdb/pdf/gsp.pdf But it seems to be too detailed for beginners IMHO. -- Best regards, Valentin valentin.likoum at ncc dor volga dot ru ------------------------------ Date: 29 May 2007 12:41:45 -0700 From: Hein RMS van den Heuvel Subject: Re: RMS_TUNE_CHECK - ROOT: Primary key index root level is high Message-ID: <1180467705.653919.5510@g4g2000hsf.googlegroups.com> On May 29, 10:18 am, Peter Weaver wrote: > For the past few months I have been working with a client who had > severe performance problems on a 4 node/14 CPU Alpha cluster. At the > start I could put one of the machines into high Interrupt and MP Sync > by doing a simple DIR/DATE on four files, each file took 6 to 8 > seconds to display. We have performance at an acceptable level now but > now I am trying to get the machines to show what they can really do. > To help keep things under control I wrote up a little DCL to run > Hein's RMS_TUNE_CHECK on a few key files and mail off a HTML formatted > message showing the results so the programmers would know when to do > converts. > > The customer asked me to help them interpret the various messages. But > I am not sure what to do with "ROOT: Primary key index root level is > high: 3 (goal=2)." Are there any settings in the FDL I can play with > that will fix this? > > Peter Weaverwww.weaverconsulting.ca > CHARON-VAX CHARON-AXP DataStream Reflection PreciseMail HP Commercial > Hardware Read the source! The 'ROOT' check is relatively simple minded. It's just a quick range check for the level given a certain file size. Just a hint. Not an edict. Send me an Email and we can discuss details. Hope this helps some, Hein van den Heuvel (at gmail dot com) HvdH Performance Consulting ------------------------------ Date: Tue, 29 May 2007 16:13:52 -0400 From: JF Mezei Subject: Re: RMS_TUNE_CHECK - ROOT: Primary key index root level is high Message-ID: <2c341$465c89a4$cef8887a$10323@TEKSAVVY.COM> Peter Weaver wrote: > The customer asked me to help them interpret the various messages. But > I am not sure what to do with "ROOT: Primary key index root level is > high: 3 (goal=2)." Are there any settings in the FDL I can play with > that will fix this? I am sure that if I am wrong, Hein and many others will pounce on me like a lynx on a mouse. But... When you EDIT/FDL the file, when you get to the Key 0 graphs type selection, you can choose a bucket size vs Index Depth. (you can define an emphasis of "flatter" with the EM keyword when in that graph mode (the one where to type FD to finish design). Then, later on, you are presented with a bucket size selection decision, you are given some suggested bucket sizes and the index depth that will be needed for those. When you look at the resulting FDL output, there doesn't seem to be any specification of index depth. My guess is that it is a matter of bucket splits in the index. Say a index bucket can contain 25 key-bucketVBN pairs. root bucket level 0: points to 25 level 1 index buckets. level 1 buckets 1 to 25 each contain pointers to 25 data buckets. So you can have a max of 625 data buckets. When you need to add the 626th data bucket, it requires a bucket split in level 1, which results in 26 index buckets at level 1. When you add that 26th index bucket at level 0, you need to add a new pointer in the root index bucket. But that one is full and needs to be split. But you cannot have 2 root buckets, so a new root is created above it to point to those 2 buckets. so the new root has 1 bucket with 2 entries. the new level 1 has 2 buckets with 26 entries the new level 2 has 26 buckets with 626 entries (pointing to data buckets) It appears that given fixed: key size, record length, and number of records to be loaded, then the bucket size is the variable that will then incluence the number of index levels. If you look in the freeware, Hein has a powerpoint presentation that provides insight in indexed file tuning. Bigger buckets require more buffer space. Global buffers let you "cache" an index. But bigger buckets mean fewer bucksts can fit in the global buffers. ------------------------------ Date: 29 May 2007 14:59:45 -0700 From: Peter Weaver Subject: Re: RMS_TUNE_CHECK - ROOT: Primary key index root level is high Message-ID: <1180475984.930063.150990@o5g2000hsb.googlegroups.com> JF Mezei wrote: >... Thanks for the tutorial; that helps. I always left the FDL editor to the programmers so I never played with it very often. > If you look in the freeware, Hein has a powerpoint presentation that > provides insight in indexed file tuning. Yes, Hein's presentations were a big help in getting this customer to the point they are at now. I talked to Hein on the phone a few minutes ago and he was very helpful and full of advice. I would like to also add that Keith Parris' various notes on locks was another big help. Keith was kind enough to answer a few emails back when I was attacking this problem from a locking point of view. (This system was really like an onion, I would find a bunch of clues that said X was a problem, fix that problem then find another problem on the layer below). T4 and Friends should be on every system. Actually I was not able to run T4 since the customer was too concerned about it impacting the system, I could not even turn on process level accounting :(. But one of the previous consultants who tried to fix this problem ran T4 for a few days last year and left the data files on the disks so I was able to work with them using TLVIS on a PC. People like Keith and Hein who publish their presentations and people like JF who a are willing to create a set of instructions and people like the T4 folks who released a great tool really make VMS a great environment to work in. Peter Weaver www.weaverconsulting.ca CHARON-VAX CHARON-AXP DataStream Reflection PreciseMail HP Commercial Hardware ------------------------------ Date: 29 May 2007 14:40:07 -0500 From: Kilgallen@SpamCop.net (Larry Kilgallen) Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: In article , Rob Brown writes: > I can't see the use of it. When would the financial guys like to use > floating point? When do they ever accept x+y=x for y<>0? When doing foreign currency trading (for various views of "foreign"). ------------------------------ Date: Tue, 29 May 2007 20:12:39 GMT From: Rob Brown Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: On Tue, 29 May 2007, Larry Kilgallen wrote: > In article , Rob Brown writes: > >> I can't see the use of it. When would the financial guys like to >> use floating point? When do they ever accept x+y=x for y<>0? > > When doing foreign currency trading (for various views of > "foreign"). This could be done just as easily using fixed point. What is the advantage of using floating point? -- Rob Brown b r o w n a t g m c l d o t c o m G. Michaels Consulting Ltd. (780)438-9343 (voice) Edmonton (780)437-3367 (FAX) http://gmcl.com/ ------------------------------ Date: Tue, 29 May 2007 17:10:20 -0400 From: JF Mezei Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: <183e4$465c96e0$cef8887a$10365@TEKSAVVY.COM> Larry Kilgallen wrote: >> I can't see the use of it. When would the financial guys like to use >> floating point? When do they ever accept x+y=x for y<>0? > > When doing foreign currency trading (for various views of "foreign"). Foreign currency trading, such as when done through SWIFT is done as fixed decimal with a controlled number of decimals everyone uses. ------------------------------ Date: Tue, 29 May 2007 17:43:37 -0400 From: JF Mezei Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: http://www.theinquirer.net/default.aspx?article=39918 Good article on the new Power6 chip with references to Alpha and recent trends in computers. Note: this page (like many pages on the inquirer) will crash Mozilla. I had to access it on Mosaic. ------------------------------ Date: Tue, 29 May 2007 19:31:25 -0700 From: "Tom Linden" Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: On Tue, 29 May 2007 13:12:39 -0700, Rob Brown wrot= e: > On Tue, 29 May 2007, Larry Kilgallen wrote: > >> In article ,= = >> Rob Brown writes: >> >>> I can't see the use of it. When would the financial guys like to us= e = >>> floating point? When do they ever accept x+y=3Dx for y<>0? >> >> When doing foreign currency trading (for various views of "foreign").= > > This could be done just as easily using fixed point. What is the = > advantage of using floating point? > The floating point referred to here is decimal floating point, not binar= y, and as such it is essentially scaled fixed decimal with a 2 digit expone= nt -- = Using Opera's revolutionary e-mail client: http://www.opera.com/mail/ ------------------------------ Date: Wed, 30 May 2007 04:00:27 GMT From: Rob Brown Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: On Tue, 29 May 2007, Tom Linden wrote: > On Tue, 29 May 2007 13:12:39 -0700, Rob Brown > wrote: > >> On Tue, 29 May 2007, Larry Kilgallen wrote: >> >>> In article >>> , Rob >>> Brown writes: >>> >>>> I can't see the use of it. When would the financial guys like to >>>> use floating point? When do they ever accept x+y=x for y<>0? >>> >>> When doing foreign currency trading (for various views of >>> "foreign"). >> >> This could be done just as easily using fixed point. What is the >> advantage of using floating point? >> > The floating point referred to here is decimal floating point, not > binary, and as such it is essentially scaled fixed decimal with a 2 > digit exponent Sure, but why use floating point if you are financial guy? I thought those guys wanted exact answers. But with floating point, if x>>y and y<>0 then x+y=x. So what is the advantage of using floating point in the financial environment? -- Rob Brown b r o w n a t g m c l d o t c o m G. Michaels Consulting Ltd. (780)438-9343 (voice) Edmonton (780)437-3367 (FAX) http://gmcl.com/ ------------------------------ Date: Wed, 30 May 2007 00:13:42 -0400 From: JF Mezei Subject: Re: World's Fastest Processor? For Now, Anyway Message-ID: <2a9b$465cfa1e$cef8887a$2144@TEKSAVVY.COM> Rob Brown wrote: > Sure, but why use floating point if you are financial guy? I thought > those guys wanted exact answers. But with floating point, if x>>y and > y<>0 then x+y=x. Without it, you use integers. if you have 53.12345 , it is probably stored as 5312345 with logic that knows that this integer is the real number multiplied by 100000. If you wish to add 34.95 to it, you need to multiply that number to match the radix of the first one (say 34.95 is stored as 3495, you need to multiply it by 1000 instead of 100000 , and then add to 5312345 and when you output it, you use a formating routine that adds the decimal point 5 digits from the right. If you have decimal floating point, you can have 53.12345 and add it directly to 34.94 and the end result can then be directtly formatted. In your logic, you can also compare it against say 100.0 In other words, it is integer arithmetic with automatic and aligned decimal point placement. Just what the financial folks need. ------------------------------ End of INFO-VAX 2007.293 ************************