INFO-VAX Thu, 24 Jul 2008 Volume 2008 : Issue 410 Contents: Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Re: VAX w/zero votes and no QuorumDisk access? Were can I buy a cheap vaxstation Re: Were can I buy a cheap vaxstation Re: Were can I buy a cheap vaxstation Re: Were can I buy a cheap vaxstation Re: Were can I buy a cheap vaxstation Re: Were can I buy a cheap vaxstation Re: Were can I buy a cheap vaxstation ---------------------------------------------------------------------- Date: Wed, 23 Jul 2008 14:05:42 -0400 From: JF Mezei Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: <48877351$0$14317$c3e8da3@news.astraweb.com> Maybe my brain has switched off portions in order to conserver energy but... What if you have a 2 node cluster with a shared common quorum disk. Each node gets 1 vote, the quorum disk gets 1 vote. Node 1 goes down, node2 can continue to work. Node2 goes down and node1 can continue to work. But what happens if the ethernet link between 1 and 2 goes down ? Node1 will think node2 has gone down, but will continue to work because it still has access to the quorum disk. Node2 will think node1 is down, but will continue to work because it still has access to the quorum disk. So you end up with a partitioned cluster with 2 nodes writing to the same quorum disk without any shared locking. Did I get this right ? Or is there some other mechanism that would force one of the 2 nodes to stop ? ------------------------------ Date: Wed, 23 Jul 2008 14:24:22 -0400 From: JF Mezei Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: <48877794$0$18540$c3e8da3@news.astraweb.com> norm.raphael@metso.com wrote: > disk. What I was asking is: Can I then boot a VAX with a common > cluster-interconnect into this cluster with zero boots despite the fact > that the VAX cannot see the quorum disk directly, The biggest issue you will have to contend with in quorum. If the alpha side quorum is 2 votes with expected votes set to 3 (node1=1 node2=1 qdisk=1) When the VAX boots, if it contributes 0 votes, when it joins the cluster it would only see 2 votes (the node votes) since it wouldn't calculate the quorum disk. So if any of the 2 nodes goes down, then the vax would go down. NOW, **if** MSCP served quorum disks are technically not blocked, then it may be different. Node3 will choose one path to the disk via MSCP (say with node1). If node1 goes down, then node3 would automatically failover to using node2 to reach the quorum disk. Back in the days of disks being attached to only one node, mscp served quorum disks didn't add any functionality. But with shared disks, I think that MSCP served ones can provide some functionality even if it doesn't add protection against cluster partitioning. ------------------------------ Date: Wed, 23 Jul 2008 18:34:29 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: JF Mezei writes: >What if you have a 2 node cluster with a shared common quorum disk. Each >node gets 1 vote, the quorum disk gets 1 vote. >Node 1 goes down, node2 can continue to work. Node2 goes down and node1 >can continue to work. >But what happens if the ethernet link between 1 and 2 goes down ? >Node1 will think node2 has gone down, but will continue to work because >it still has access to the quorum disk. Node2 will think node1 is down, >but will continue to work because it still has access to the quorum disk. >So you end up with a partitioned cluster with 2 nodes writing to the >same quorum disk without any shared locking. >Did I get this right ? Or is there some other mechanism that would force >one of the 2 nodes to stop ? That's part of the purpose of a quorum disk. The nodes will detect each other through the data in the quorum file, and one will bugcheck. (did you think the quorum file was write-only or something?) ------------------------------ Date: Wed, 23 Jul 2008 19:11:19 +0000 (UTC) From: moroney@world.std.spaamtrap.com (Michael Moroney) Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: norm.raphael@metso.com writes: >All that is fascinating, but does not relate to my situation. >I (will) have a cluster with 2 AS systems and a quorum disk each >with one vote. The disks are on SAN=Fibrechannel, including the quorum >disk. What I was asking is: Can I then boot a VAX with a common >cluster-interconnect into this cluster with zero boots despite the fact >that the VAX cannot see the quorum disk directly, and in that case, >should the VAX be sysgened with or without the quorum disk - I now >believe without. This is very similar to the old voteless satellite workstation setup. The difference is the satellites usually booted over the network (their system disk was on a "big" VAX, not local) and often had no disks at all, and you almost certainly have a local disk for your voteless VAX. The satellites were configured with 0 votes and no quorum disk, regardless of the situation on the "big" nodes. Nodes should not be configured with a quorum disk unless they have direct (not MSCP) access to it. Clustering is, however, smart enough to not use MSCP for the quorum disk if it's only reachable that way. Why not skip the quorum disk and give all 3 nodes a vote? ------------------------------ Date: Wed, 23 Jul 2008 19:33:39 GMT From: John Santos Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: In article <48877794$0$18540$c3e8da3@news.astraweb.com>, jfmezei.spamnot@vaxination.ca says... > norm.raphael@metso.com wrote: > > > disk. What I was asking is: Can I then boot a VAX with a common > > cluster-interconnect into this cluster with zero boots despite the fact > > that the VAX cannot see the quorum disk directly, > > The biggest issue you will have to contend with in quorum. If the alpha > side quorum is 2 votes with expected votes set to 3 (node1=1 node2=1 > qdisk=1) > > When the VAX boots, if it contributes 0 votes, when it joins the cluster > it would only see 2 votes (the node votes) since it wouldn't calculate > the quorum disk. So if any of the 2 nodes goes down, then the vax would > go down. I'm pretty sure this is wrong. The way I think it work is that one of the systems that can see the quorum disk functions as a quorum disk watcher (quorum disk server?), and supplies both it's own votes and the QD votes to the nodes that can't see the quorum disk. So the VAX would see all three votes. If the QD watcher goes down, one of the other nodes (in this case, the other Alpha) that can still see the quorum disk takes over the role of quorum disk watcher and continues to supply the QD votes to the cluster. All this happens (unless you've misconfigured) in less than 2 RECNXINTERVAL's, so the cluster will reconfigure itself and keep going. > > NOW, **if** MSCP served quorum disks are technically not blocked, then > it may be different. Node3 will choose one path to the disk via MSCP > (say with node1). If node1 goes down, then node3 would automatically > failover to using node2 to reach the quorum disk. > MSCP served quorum disks *are* blocked. (You can serve them, so other nodes can get at regular files on them just like non-quorum disks, but systems without direct access can't be quorum disk watchers.) > Back in the days of disks being attached to only one node, mscp served > quorum disks didn't add any functionality. But with shared disks, I > think that MSCP served ones can provide some functionality even if it > doesn't add protection against cluster partitioning. > What days of disks being attached to only one node? You mean before the VAX-11/780+VAX-11/785+dual-ported RP06 system/quorum disk that was the 1st VAXCluster I installed at a customer site? (Possible memory failure: it might have been an RP05!) There's no point in MSCP-served quorum disks, since the MSCP server(s) have to be up for the MSCP clients to see the disk, so the MSCP server might just as well have the votes or, if there's more than one of them, use it for a quorum disk amongst themselves, excluding the clients. If the MSCP server goes down, and it's serving the quorum disk votes, everything knows right away, but if each client were trying to access the QD, there could be long delays while they all look around for an alternate MSCP server, extending the cluster state transition needlessly. -- John ------------------------------ Date: Wed, 23 Jul 2008 19:49:30 GMT From: John Santos Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: In article <8266ff18-1181-43b2-9412-b06feacec0cb@ 26g2000hsk.googlegroups.com>, hvlems@freenet.de says... > On 21 jul, 15:05, norm.raph...@metso.com wrote: > > If I have a cluster w/2 AS systems and a quorum disk (each with one vote, > > quorum=2), > > Can I boot a vax into this cluster (with zero votes) if it has no direct > > access to the quorum disk? > > If yes, would it need the quorum disk sysgen parameters set and that disk > > mspc-served from > > one or both of the AS systems? > > The documentation tells us that a quorum disk must be directly > attached to a system, if you want to use the quorum disk sysgen > parameter. > Directly attached means DSSI, SCSI or CI (and possibly fiber channel > but I have no experience with that). It may not be a network served > disk in an LAVc. Also dual-ported massbus disks, and (possibly) dual-ported SDI disks on KDA50, UDA50 and KDM70 controllers. (I'm not sure about the SDI disks since I tried to set up a hobbyist cluster this way (a pair of Qbus MicroVAXes each with a KDA50 and a dual-ported RA72), and couldn't get it to work, but I was fighting bad SDI cables.) Can Infoservers serve a quorum disk over ethernet? (They use LAST protocol, not MSCP, IIRC.) Also, I think someone wrote a freeware PC-based disk server that *may* work by emulating an Infoserver (I think.) DAGS. > Furthermore, there may be just one quorum disk in a cluster. > So in your example, when both AS systems have a quorum disk then the > conclusion is that this quorum disk is on a common SCSI bus shared by > these two AS's. > Since a VAX cannot share a SCSI bus at all, it cannot have knowledge > about that quorum disk. > In a two node AS cluster a third participant could prevent a cluster > transition happen. That third participant could either be a quorum > disk or a quorum VAX. The disk was thought to be somewhat cheaper than > VAX. Adding a quorum disk to the VAX (and only to that system) is > technically possible but quite useless in terms of uptime because one > could just as well add the quorum disk votes to the votes of the VAX. > However in that example, the quorum disk would only e accessed by the > VAX and doesn't need to be visible to the AS systems. > > Incidentally, a two node NI cluster where each node has access to its > own quorum disk will boot and function. When both disks and both hosts > (VAX or Alpha) each have VOTES=1 and QUORUM=2 the following happens. > Both nodes boot and will recognize each other (via the network). > ANALYSE/SYSTEM shows that the quorum disks have no function in the > cluster quorum calculations. As soon as one node fails, the remaining > node suddenly remembers it has a quorum disk, uses that vote to > replce the lost one from the failed host, reaches quorum and starts to > function again. The way this works is NOT supported and DEC, errr HP, > states that damaga to on-dis structures is possible if one decides to > operate a cluster that way. For hobbyist use I find it rather > convenient though :-) > Hans > -- John ------------------------------ Date: 23 Jul 2008 15:57:07 -0500 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: In article <48877351$0$14317$c3e8da3@news.astraweb.com>, JF Mezei writes: > > So you end up with a partitioned cluster with 2 nodes writing to the > same quorum disk without any shared locking. > > Did I get this right ? Or is there some other mechanism that would force > one of the 2 nodes to stop ? If one of the nodes tries to create, read, delete, or in any other way manipulate a lock it will find out that it can't talk to the other. In the meantime there are sanity timers that are likely to expire. The lock manager and the sanity timers do not use the quorum disk for communication. ------------------------------ Date: Wed, 23 Jul 2008 17:52:35 -0400 From: norm.raphael@metso.com Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: This is a multipart message in MIME format. --=_alternative 00782A718525748F_= Content-Type: text/plain; charset="US-ASCII" moroney@world.std.spaamtrap.com (Michael Moroney) wrote on 07/23/2008 03:11:19 PM: > norm.raphael@metso.com writes: > > >All that is fascinating, but does not relate to my situation. > >I (will) have a cluster with 2 AS systems and a quorum disk each > >with one vote. The disks are on SAN=Fibrechannel, including the quorum > >disk. What I was asking is: Can I then boot a VAX with a common > >cluster-interconnect into this cluster with zero boots despite the fact > >that the VAX cannot see the quorum disk directly, and in that case, > >should the VAX be sysgened with or without the quorum disk - I now > >believe without. > > This is very similar to the old voteless satellite workstation setup. > The difference is the satellites usually booted over the network (their > system disk was on a "big" VAX, not local) and often had no disks at all, > and you almost certainly have a local disk for your voteless VAX. > The satellites were configured with 0 votes and no quorum disk, regardless > of the situation on the "big" nodes. > > Nodes should not be configured with a quorum disk unless they have direct > (not MSCP) access to it. Clustering is, however, smart enough to not > use MSCP for the quorum disk if it's only reachable that way. > > Why not skip the quorum disk and give all 3 nodes a vote? Michael, As I 'splained earlier in the thread, the VAX is leaving the cluster. I just want a way to reboot it into the cluster temporarily in case I have to recover/reconstruct something before I sell it. --=_alternative 00782A718525748F_= Content-Type: text/html; charset="US-ASCII"
moroney@world.std.spaamtrap.com (Michael Moroney) wrote on 07/23/2008 03:11:19 PM:

> norm.raphael@metso.com writes:
>
> >All that is fascinating, but does not relate to my situation.
> >I (will) have a cluster with 2 AS systems and a quorum disk each
> >with one vote.  The disks are on SAN=Fibrechannel, including the quorum
> >disk.  What I was asking is:  Can I then boot a VAX with a common
> >cluster-interconnect into this cluster with zero boots despite the fact
> >that the VAX cannot see the quorum disk directly, and in that case,
> >should the VAX be sysgened with or without the quorum disk - I now
> >believe without.
>
> This is very similar to the old voteless satellite workstation setup.
> The difference is the satellites usually booted over the network (their
> system disk was on a "big" VAX, not local) and often had no disks at all,
> and you almost certainly have a local disk for your voteless VAX.
> The satellites were configured with 0 votes and no quorum disk, regardless
> of the situation on the "big" nodes.
>
> Nodes should not be configured with a quorum disk unless they have direct
> (not MSCP) access to it.  Clustering is, however, smart enough to not
> use MSCP for the quorum disk if it's only reachable that way.
>
> Why not skip the quorum disk and give all 3 nodes a vote?

Michael,

As I 'splained earlier in the thread, the VAX is leaving the cluster.
I just want a way to reboot it into the cluster temporarily in case I
have to recover/reconstruct something before I sell it. --=_alternative 00782A718525748F_=-- ------------------------------ Date: Wed, 23 Jul 2008 20:45:33 -0400 From: JF Mezei Subject: Re: VAX w/zero votes and no QuorumDisk access? Message-ID: <4887d101$0$1833$c3e8da3@news.astraweb.com> norm.raphael@metso.com wrote: > As I 'splained earlier in the thread, the VAX is leaving the cluster. > I just want a way to reboot it into the cluster temporarily in case I > have to recover/reconstruct something before I sell it. based on what others have said, I would suggest the VAX be configured without cluster quorum disk. When it boots, it will see the votes offered by the other nodes and decide if there are enough votes to continue to boot into this cluster. The VAX will just see one of the alphas offering its own vote(s) and that of the quorum disk. To be sure, you could also set the vax to EXPECTED_VOTES=1 this way, years from now, when you try to boot it, it will continue to boot no matter how many votes are left on the cluster. ------------------------------ Date: Wed, 23 Jul 2008 13:37:26 -0700 (PDT) From: nolanh@ymail.com Subject: Were can I buy a cheap vaxstation Message-ID: <3ebb9366-8387-42df-be2b-e0b763d77fd4@z72g2000hsb.googlegroups.com> I currently use vms just by using telnet but I would like to have a vaxstation of my own , does anyone know were I cant get my own for a decent price. ------------------------------ Date: 23 Jul 2008 20:42:59 GMT From: VAXman- @SendSpamHere.ORG Subject: Re: Were can I buy a cheap vaxstation Message-ID: <488797d3$0$20904$607ed4bc@cv.net> In article <3ebb9366-8387-42df-be2b-e0b763d77fd4@z72g2000hsb.googlegroups.com>, nolanh@ymail.com writes: >I currently use vms just by using telnet but I would like to have a >vaxstation of my own , does anyone know were I cant get my own for a >decent price. Do you need graphics? If not, a small microVax would the trick. Contact me privately. Put "[PASS:VAXman]" in the subject to avoid the anti-SPAM filter from rejecting you. -- VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)COM ... pejorative statements of opinion are entitled to constitutional protection no matter how extreme, vituperous, or vigorously expressed they may be. (NJSC) Copr. 2008 Brian Schenkenberger. Publication of _this_ usenet article outside of usenet _must_ include its contents in its entirety including this copyright notice, disclaimer and quotations. ------------------------------ Date: 23 Jul 2008 15:59:16 -0500 From: koehler@eisner.nospam.encompasserve.org (Bob Koehler) Subject: Re: Were can I buy a cheap vaxstation Message-ID: <6C6JzAy2a9PH@eisner.encompasserve.org> In article <3ebb9366-8387-42df-be2b-e0b763d77fd4@z72g2000hsb.googlegroups.com>, nolanh@ymail.com writes: > I currently use vms just by using telnet but I would like to have a > vaxstation of my own , does anyone know were I cant get my own for a > decent price. Ebay. And many other places you can find on the Internet. Also SIMH won't cost you a thing. ------------------------------ Date: Wed, 23 Jul 2008 15:53:11 -0500 (CDT) From: sms@antinode.info (Steven M. Schweda) Subject: Re: Were can I buy a cheap vaxstation Message-ID: <08072315531186_20200492@antinode.info> From: nolanh@ymail.com > I currently use vms just by using telnet but I would like to have a > vaxstation of my own , does anyone know were I cant get my own for a > decent price. I know many places where you can't get one. Or did you wish to know where you _can_ get one? Where are you? Define "decent". Why not an Alpha, so you could use some features added since 2001? Shipping on old VAXstations tends to be higher than their value, which tends to diminish their abundance on Ebay. ------------------------------------------------------------------------ Steven M. Schweda sms@antinode-info 382 South Warwick Street (+1) 651-699-9818 Saint Paul MN 55105-2547 ------------------------------ Date: Wed, 23 Jul 2008 17:26:29 -0400 From: "Richard B. Gilbert" Subject: Re: Were can I buy a cheap vaxstation Message-ID: nolanh@ymail.com wrote: > I currently use vms just by using telnet but I would like to have a > vaxstation of my own , does anyone know were I cant get my own for a > decent price. The good ones are not cheap! DEC made a VAXstation I, VAXstation II, VAXstation 2000, VAXstation 3100, VAXstation 4000/VLC, 4000-60, 4000-90, 4000/90A. That list is, more or less, in increasing order of desirability and cost. The newest of them is at least nine or ten years old now. They were good machines in their day but their day is long gone! If I've missed a few models, I'm sure someone will pop up and name them! Some of these may not be supported under current versions of VMS!! I would keep an eye on e-Bay, Craig's List, and other "for sale" sites and lists. ------------------------------ Date: Wed, 23 Jul 2008 15:25:40 -0700 (PDT) From: Rich Jordan Subject: Re: Were can I buy a cheap vaxstation Message-ID: <976309ac-11cb-4fd0-809d-9e748b785083@z66g2000hsc.googlegroups.com> On Jul 23, 4:26=A0pm, "Richard B. Gilbert" wrote: > nol...@ymail.com wrote: > > I currently use vms just by using telnet but I would like to have a > > vaxstation of my own , does anyone know were I cant get my own for a > > decent price. > > The good ones are not cheap! =A0DEC made a VAXstation I, VAXstation II, > VAXstation =A02000, VAXstation 3100, VAXstation 4000/VLC, 4000-60, > 4000-90, 4000/90A. =A0That list is, more or less, in increasing order of > desirability and cost. =A0The newest of them is at least nine or ten year= s > old now. =A0They were good machines in their day but their day is long > gone! =A0If I've missed a few models, I'm sure someone will pop up and > name them! =A0Some of these may not be supported under current versions o= f > VMS!! > > I would keep an eye on e-Bay, Craig's List, and other "for sale" sites > and lists. The most recent VAXstation was the 4000/96. Two of those sold on Ebay recently for ~$700 each if I recall. Slower machines tend to be much cheaper, but watch the shipping (as Steven mentioned); they are very solid, well built boxes and weigh a bit. The 4000/90/90a/96 series are still priced at painful levels for an unfunded home hobbyist (IMO). The 4000/60 is a decent box that can probably be had about as cheap as any of the slower ones and can handle 104MB RAM if I remember correctly (boss made my pack up all my old SOCs... sigh...). The 4000s also have another benefit; if you're lucky you might find one with (or find a separate) Nemonix enhancement board; these were available with cache and performance enhancements, Ultra-SCSI, and 100Base-T support, capabilities not available on any earlier workstation. The Nemonix boards show up on Ebay very occasionally. The 4000 VLC is very limited on memory (24MB) and internal storage. The 3100/m76 is hard to get memory for (it was unique) so if you find one with less than maximum (32MB) memory, consider that a problem. If you settle for any 3100 model, try to find an SPX 8-plane color equipped unit if possible. The SPX was the fastest graphics adapter by far and even for text (terminal window) use made the system much more enjoyable. You will need a finicky monitor to work with a GPX color or monochrome VAXstation; the 1024x864 fixed frequency output with sync-on-green will throw most PC monitors for a loop. However with an SPX card you can run 1280x1024 (still sync on green), and many older flat panels support that as a native resolution at 60Hz refresh (which the SPX supports); I have a NEC Multisync LDC2010 connected to my desktop 3100/30 SPX with the factory cable and it works just fine. The 3100/ m76 SPX is connected to a genuine DEC VRC 20" monitor and is also fine. You will also need the correct LK series keyboard and a compliant mouse; these are not PC compatible components. Best if they come with whatever system you buy... Newer flat panels have almost all gone to cheap crap nontraditional "widescreen" resolutions and dropped BNC connectors and sync-on-green support. You need to check the specs. Sorry, our VAXstations are not spares; they're just too darn useful and we like the fact that the oldest is 19+ years old and going strong. =3D=3D=3D=3D=3D=3D=3D=3D As others have mentioned, unless you specifically want to do VAX programming (not just VMS), you might seriously consider an Alphastation instead. The older models with PCI slots (AlphaStation 200-4/166 and 4/233, model 255/233 and 255/300, and Personal Workstations (433 - 600MHz, look for an 'au' model) can all be had relatively inexpensively (make sure they come with VMS compatible graphics cards though), use standard peecee monitors and PS2 mice, and in a pinch can even use a crappy peecee (PS2) keyboard until you get a real VMS keyboard. Probably keep away from the 2000 and 3000 series Alphas which are far more limited or far bigger/heavier. A DS10 would be stellar but they tend to be pricey still; stay clear of the DS10-L, with only one slot if you want both decent (SCSI) storage and graphics you have to get a very expensive combo board. Note that the older units (AS200/AS255 are not recommended if you're thinking of running a browser or working with such multimedia capabilities as VMS provides; they're too old and the EV4 processors lacked some instructions that made handling unaligned and smaller data items far more efficient on later models. If you just want a multi- text-window VMS terminal box for programming and playing, though, they're just fine. I still use my AS200 4/233 at work quite a bit for those purposes. Rich ------------------------------ Date: Wed, 23 Jul 2008 19:53:43 -0400 From: JF Mezei Subject: Re: Were can I buy a cheap vaxstation Message-ID: <4887c4c5$0$18540$c3e8da3@news.astraweb.com> nolanh@ymail.com wrote: > I currently use vms just by using telnet but I would like to have a > vaxstation of my own , does anyone know were I cant get my own for a > decent price. You need to specify where you are. Shipping costs can be non trivial. Another alternative is to contact David at islandco.com (he posts here frequently) and ask him about the DS10L boxes with a graphics card and a 30 gig disk + shipping by fedex to you. This is an alpha and you can get more modern software on alpha because VAX was unceremoniously abandonned before promised releases. ------------------------------ End of INFO-VAX 2008.410 ************************