Re: White paper: CHARON-VAX cluster w/ shared SCSI disks
From: konabear (maurert_at_ameritech.net)
Date: 10/17/03
- Next message: Wilm Boerhout: "Re: White paper: CHARON-VAX cluster w/ shared SCSI disks"
- Previous message: JF Mezei: "Re: Cost of process creation on Unix"
- In reply to: Wilm Boerhout: "Re: White paper: CHARON-VAX cluster w/ shared SCSI disks"
- Next in thread: Michael Moroney: "Re: White paper: CHARON-VAX cluster w/ shared SCSI disks"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Fri, 17 Oct 2003 18:26:36 GMT
That and you can contract OpenVMS support for your CharonVAX. HP will
support VMS on a CharonVAX provided it's running on HP hardware. The only
caveat is that should VMS exhibit an unexpected behavior, the customer may
be ask to duplicate the problem on a VAX.
What I'm not sure of is if Engineering would step up to a problem on the VAX
Shared SCSI cluster mentioned here. But as always, many things in OpenVMS
work even if they aren't supported. Just don't expect OpenVMS Engineering
to fix it if it breaks.
The solutoin in the white paper it great, but would who support the shared
SCSI if there is a problem?
Also being on the team that support OpenVMS clusters at the CSC, I STRONGLY
recommend that the instructions include setting EXPECTED_VOTES to 3 on both
systems. If left at the defaults of 1 then a partitioned cluster would be
possible. I had a partitioned cluster happen the other day. We could get
into a rathole quickly here.
Since you recommend configing with a quorum disk, I admit it is unlikely
that a partitioned cluster would happen. If NodeA booted while Ethernet
were unavailable for SCS, NodeA would still see activity in the quorum file
to let it know NodeB is using the system disk. Not seeing NodeB via SCS
NodeA should refuse to boot. The danger would be a down Quorum disk .AND.
broken Ethernet. But the point is EXPECTED_VOTES is the design feature to
protect the cluster and the data. The white paper does a good job of
warning about boot sequence during setup to avoid partitioning, so I know
avoiding partitioning is important to the author.
Todd
Todd
"Wilm Boerhout" <wboerhoutOLD@PAINTvxcompany.com> wrote in message
news:bmh9oq$fei$1@reader08.wxs.nl...
>
>
> H Vlems wrote:
>
> > Whatever way you may want to put it,
> > for VMS system managers the Charon setup is a straight forward NI
cluster.
> > Which incidentally may also be achieved if you run simh.
>
>
> You may have seen by now that it is not an NI-(LAVC)-cluster at all.
>
> As for comparing it to simh, that's not fair at all. simh (I've used it
> and the public domain / open source predecessors for a long time. Used
> to load RSTS/E on the PDP-11 simulator just to run Adventure)
>
> simh is perfectly suitable for the hobbyist and educational environment.
> I bet it can win over some new VMS enthusiasts.
>
> CHARON-VAX on the other hand is production class software. One of our
> customers runs it with 800 users logged in and heavy batches at night on
> a single soon to be clustererd Compaq Proliant DL360 / DL380, replacing
> a VAX 7000-620.
>
> The cluster that led to writing the white paper delivers 100 VUPS! Be
> careful, children, this is not a toy :-)
>
> --
> Wilm Boerhout
> VX Company, The Netherlands
>
> wboerhoutOLD@PAINTvxcompany.com
> (remove OLD PAINT from reply address)
>
> *** my opinions are strictly my own ***
>
>
- Next message: Wilm Boerhout: "Re: White paper: CHARON-VAX cluster w/ shared SCSI disks"
- Previous message: JF Mezei: "Re: Cost of process creation on Unix"
- In reply to: Wilm Boerhout: "Re: White paper: CHARON-VAX cluster w/ shared SCSI disks"
- Next in thread: Michael Moroney: "Re: White paper: CHARON-VAX cluster w/ shared SCSI disks"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
