INFO-VAX Tue, 02 Jan 2007 Volume 2007 : Issue 4 Contents: Re: Alpha console ANSI colour support ? Re: DECterm special escape sequences ? Re: Getting cluster to ignore temporary loss of ethernet ? Re: Getting cluster to ignore temporary loss of ethernet ? Re: Renaming a node in cluster Re: Renaming a node in cluster Re: Renaming a node in cluster Re: Renaming a node in cluster Re: Renaming a node in cluster Re: Setting sound frequency for BEEPs on DS10L Re: Setting sound frequency for BEEPs on DS10L Re: Status of DECNOTES ? Re: Status of DECNOTES ? Re: Status of DECNOTES ? Re: UnZip 5.53 VMS and bzip2 ---------------------------------------------------------------------- Date: 2 Jan 2007 07:13:51 -0800 From: "Ian Miller" Subject: Re: Alpha console ANSI colour support ? Message-ID: <1167750831.636925.203300@h40g2000cwb.googlegroups.com> last time I played, that console was around VT52 level of capability i.e. no colour. The output of the system startup can be recorded to a file. See SYSMAN STARTUP SET OPTION/OUTPUT=FILE or just set STARTUP_P2 to D ------------------------------ Date: 2 Jan 2007 07:52:08 -0600 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: DECterm special escape sequences ? Message-ID: In article , Paul Sture writes: > > "VT100 > Digitalšs first ANSI-compliant terminal, introduced in August 1978." > > which tallies extremely well with my memories of a PDP upgrade and > moving that system out of what could best be described as a broom > cupboard into a proper computer room, our first VT100 included. When we got our first VAX, in 1980, we were able to buy VT100 clones already. ------------------------------ Date: Tue, 2 Jan 2007 14:37:54 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: Getting cluster to ignore temporary loss of ethernet ? Message-ID: JF Mezei writes: >Malcolm Dunnett wrote: >> A large value for RECNXINTERVAL is the simplest way to address this. The >> downside is that it will make a legitimate cluster transition ( eg when one >> node crashes ) take longer. >What happens if you have a RECNXINTERVAL set to say 5 minutes, a node >crashes and reboots within that time period ? what will the rest of the >cluster say when that node asks to join the cluster when the rest of the >cluster still thinks that node is alread in the cluster ? Good question. Let us know :-) I do know RECNXINTERVAL is supposed to be short enough so that the cluster kicks out the node before it gets a chance to reboot to the point where it can do damage by having two instances of something. >In terms of my own needs, since this appears to be a dynamic sysgen >parameter, I guess when I plan some system maintenance, I can use SYSMAN to >execute a sysgen script to change the active value of RECNXINTERVAL to some >high limit, and then once the switch maintenance is done, does it back to >its normal limit. I've done this many times. ------------------------------ Date: 2 Jan 2007 07:15:54 -0800 From: "Volker Halle" Subject: Re: Getting cluster to ignore temporary loss of ethernet ? Message-ID: <1167750953.992758.136190@s34g2000cwa.googlegroups.com> JF, temporarily setting RECNXINTERVAL to a higher value cluster-wide is the way to go. An unusal high value of RECNXINTERVAL is NOT a problem, if a node crashes or is shut down. During the crash or shutdown, the node will send a 'last gasp' datagram to all other cluster members, so it will be removed from the cluster immediately. You only get a long-lasting state transition, if the node stops sending cluster hello messages without sending a 'last gasp' datagram, i.e. if you HALT the node or just unplug the network cable. What happens, if a node reboots during a network outage, mainly depends on the votes/quorum settings of that node and the other nodes in the cluster. When the node boots and can't see the other cluster members, it may create it's own cluster (assuming it has enough votes to do so) with a new INCARNATION number. When the network connectivity is restored, the cluster cannot be re-established if the node has a new incarnation number, so some nodes need to bugcheck with a CLUEXIT crash to be able to 're-join' the cluster. Again, it depends on votes,quorum and SCSSYSTEMID settings, which nodes have to reboot. You can use a second LAN interface on the VAXes in the same way as on Alpha or Itanium systems. PEDRIVER will start the SCS protocol on each network adapter it finds. Trying to HALT and CONTINUE the systems will most likely not work as expected, especially for SMP systems. Volker. ------------------------------ Date: 2 Jan 2007 04:36:01 -0800 From: "Ian Miller" Subject: Re: Renaming a node in cluster Message-ID: <1167741361.200856.91330@i12g2000cwa.googlegroups.com> you have to change name and scssystemid otherwise the cluster will remember, ------------------------------ Date: Tue, 02 Jan 2007 07:53:19 -0500 From: JF Mezei Subject: Re: Renaming a node in cluster Message-ID: Ian Miller wrote: > you have to change name and scssystemid otherwise the cluster will > remember, What if I bring down 1028-CRAP, keep it down, then do sequential reboot of the other 3 nodes, and then bring 1028-NODE4 up ? Right now, my oldest node remembers, in SHOW CLU/CONT, my VAX workstation BIKE1. Interestingly the other VAX which acted as a boot node show BIKE1 as "VMS BOOT". But neither of the alphas know about BIKE1 since they have since rebooted a few times. based on the output of SHOW CLUSTER, it would appear that information about departed nodes doesn't get propagated to nodes as they join the cluster. If SHOW CLU/CONT shows the true view of the cluster from the point of view of the node it runs on, then it would lead me to believe that new nodes do not get told about dead nodes. ------------------------------ Date: 2 Jan 2007 05:42:51 -0800 From: etmsreec@yahoo.co.uk Subject: Re: Renaming a node in cluster Message-ID: <1167745371.635376.229320@v33g2000cwv.googlegroups.com> No JF, that won't work. As Ian and Robert have both said, you need to reboot the _cluster_ if you change one and not the other. If 1028 changes its nodename only, you have to reboot the cluster. If CRAP changes its SCSSYSTEMID only, you have to reboot the _cluster_. Only if you change BOTH or NEITHER will the node be readmissable into the cluster. It's an all or nothing thing. Steve JF Mezei wrote: > Ian Miller wrote: > > you have to change name and scssystemid otherwise the cluster will > > remember, > > What if I bring down 1028-CRAP, keep it down, then do sequential reboot of > the other 3 nodes, and then bring 1028-NODE4 up ? > > Right now, my oldest node remembers, in SHOW CLU/CONT, my VAX workstation > BIKE1. Interestingly the other VAX which acted as a boot node show BIKE1 as > "VMS BOOT". But neither of the alphas know about BIKE1 since they have > since rebooted a few times. > > > based on the output of SHOW CLUSTER, it would appear that information about > departed nodes doesn't get propagated to nodes as they join the cluster. > > > If SHOW CLU/CONT shows the true view of the cluster from the point of view > of the node it runs on, then it would lead me to believe that new nodes do > not get told about dead nodes. ------------------------------ Date: 2 Jan 2007 08:13:14 -0600 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: Renaming a node in cluster Message-ID: In article , JF Mezei writes: > Ian Miller wrote: >> you have to change name and scssystemid otherwise the cluster will >> remember, > > What if I bring down 1028-CRAP, keep it down, then do sequential reboot of > the other 3 nodes, and then bring 1028-NODE4 up ? As long as any node of the cluster is up, the cluster remembers the old use of the same SCSSYSTEMID. Only by shutting down the entire cluster at the same time is that knowedge lost. The only workaround is to change the SCSSYSTEMID. Been there, done that; had one system with two boot roots with different SCSNODE and SCSSYSTEMID. We didn't have an UPS so on the next power hit the system came up with te default root and all knowledge of the original SCSNODE/SCSSYSTEMID combination was lost, then I just cleaned up the spare root. ------------------------------ Date: Tue, 2 Jan 2007 14:48:28 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: Renaming a node in cluster Message-ID: JF Mezei writes: >1025 NODE1 1 vote (oldest node generally granting cluster membership) >1026 NODE2 2 votes (changing to 1 vote) >1027 NODE3 1 vote >1028 CRAP 0 vote >Ok, so now I want to rename "CRAP" to NODE4, keep its same SCSSYSTEMID and >give it 1 vote. And I want this without having to shutdown the whole >cluster (aka: wish to preserve cluster uptime :-) >If I update CRAP to become NODE4, and then reboot it, NODE4 should be >refused membership in the cluster and be prevented from continued booting, >correct ? Correct. >If I then sequentially reboot each of the other 3 nodes, would they not >forget about 1028-CRAP and eventually accept 1028-NODE4 into the cluster ? They do. But as long as one remembers the old value, it'll be refused. >If 1028-NODE4 has been refused entry into the cluster, will it be waiting >or going back to the >>> ? I don't really remember, but I think it waits, and keeps trying to join the cluster. >If the above is expected to work, would the order in which I reboot the 3 >nodes matter ? (eg: oldest node generally is the veteran that has >responsability to accept/deny cluster membership). Order doesn't matter. Neither does the identity of the node whose name you see upon joining the cluster. I don't remember how that node is determined, whether it's the lowest scssystemid, the node the new member sees first, the node that sees the new member first, the oldest, or semi-random (the first CSB or CLUB block on some queue). Cluster nodes are considered equals, except for votes. ------------------------------ Date: Tue, 02 Jan 2007 13:07:46 GMT From: lederman@star.enet.dec.DISABLE-JUNK-EMAIL.com (Bart Z. Lederman) Subject: Re: Setting sound frequency for BEEPs on DS10L Message-ID: Try using xset, which can be found in DECW$UTILS You can put it into your login.com file if you want the tone set automatically. Note that what you can do with the built-in sound on a DS10x may be rather limited. Bart ------------------------------ Date: Tue, 2 Jan 2007 14:22:25 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: Setting sound frequency for BEEPs on DS10L Message-ID: lederman@star.enet.dec.DISABLE-JUNK-EMAIL.com (Bart Z. Lederman) writes: >Try using xset, which can be found in DECW$UTILS >You can put it into your login.com file if you want the tone >set automatically. Or, click on the Style Manager icon (fourth from right on the task bar), click on Beep, change tone and duration, save settings, log out of DECwindows (or update the DECwindows parameter files, if you know how). The settings will take place when you log in again. ------------------------------ Date: 2 Jan 2007 03:59:14 -0800 From: "Ian Miller" Subject: Re: Status of DECNOTES ? Message-ID: <1167739154.620033.216550@v33g2000cwv.googlegroups.com> NOTES is now freeware. See http://h71000.www7.hp.com/freeware/freeware70/notes/ ------------------------------ Date: 2 Jan 2007 04:13:07 -0800 From: etmsreec@yahoo.co.uk Subject: Re: Status of DECNOTES ? Message-ID: <1167739987.512419.243740@42g2000cwt.googlegroups.com> Check the freeware JF - the last version (prior to v8.3 distribution) had Notes on it. Steve JF Mezei wrote: > Low and behold, having downgraded VMS from 7.2 to 7.3 on my remaining VAXes > so they can interoperate with Alpha 8.3, I have just found that VAX NOTES > is no longer working. > > When you type NOTES, it simplay says "File not found". > > It seems it was linked against XDPS (Display Postscript). So even when you > request the character cell interface, you can't start notes. > > So, I moved the 3 XDPS files from sys$library from a backup of 7.2 to the > 7,3 running system. Now, it doesn't complain about missing files, but > complains about %SYSTEM-F-SHRIDMISMAT, ident mismatch with shareable image > > :-( :-( :-( > > I look at my 7.3 VAX condist and cannot find NOTES listed in it. (could it > be burried inside the ALLIN1 kits ?) > > > I was aware that DECWRITE was recompiled to exclude support for XDPS (as > well as many utilities part of VMS). But had not heard about NOTES and if > NOTES had been recompiled for 7.3, shouldn't it have been included in the > condists for VAX ? (my VAX cd set dates from dec 04). > > > So, if "they" forgot to recompile NOTES, and the product is truly broken, > is there a cookbook on how to patch an image and/or shareable images to > allow the ident mismatches instead of complaining about them ? ------------------------------ Date: 2 Jan 2007 05:37:47 -0800 From: etmsreec@yahoo.co.uk Subject: Re: Status of DECNOTES ? Message-ID: <1167745067.024303.163700@h40g2000cwb.googlegroups.com> As you said the other day JF, try it - isn't that what hobbyists are for? JF Mezei wrote: > Ian Miller wrote: > > NOTES is now freeware. See > > http://h71000.www7.hp.com/freeware/freeware70/notes/ > > Many thanks. > > The readme file provides some information. And I was able to use > NOTES/NODECWINDOWS on my VAX to bypass the ident mismatch (or missing files > from display postscript). > > Will installing the patched files re-enable decwidnows NOTES on VAX and/or > Alpha ? Or is the product stuck in its character cell mode ? (I haven't > brought ALL-IN-1 back up since my disk crash, but will have to check the > impact because it uses callable notes). ------------------------------ Date: Tue, 2 Jan 2007 01:54:22 -0600 (CST) From: sms@antinode.org (Steven M. Schweda) Subject: Re: UnZip 5.53 VMS and bzip2 Message-ID: <07010201542278_2020028F@antinode.org> From: "Christian Spieler" > a) Get the generic source code distribution for bzip2 (preferably the > original sources from www.bzip2.org) and extract them to the bz2lib > subdirectory of the unpacked UnZip source distribution. > b) Start the build procedure for UnZip with the following command ... > (maybe some command parameters enabling bzip2 support have to be included) One more potential problem... On IA64, until very recently (*), the C compiler had an optimization bug which affected bzip2's huffman.c, causing bad results. When building a complete bzip2 kit on VMS, it's possible to run the tests, which will reveal whether your compiler worked right. (If not, a moderately manual work-around is available.) If only the BZ_NO_STDIO library is built, you can't run the normal tests, so you can't easily see if you have the problem or not. Of course, building the complete bzip2 kit on VMS requires non-generic, non-original sources. Everything's complicated. (*) Bad: HP C V7.2-001 (on OpenVMS IA64 V8.2-1). Good: HP C V7.2-022 (on OpenVMS IA64 V8.3). ------------------------------------------------------------------------ Steven M. Schweda sms@antinode-org 382 South Warwick Street (+1) 651-699-9818 Saint Paul MN 55105-2547 ------------------------------ End of INFO-VAX 2007.004 ************************