[Simh] SIMH VAXcluster - shadow disk issues

[Stephen Hoffman]

On Fri, 20 Jan 2012 08:10:59 +0100, Lennert Van Alboom <lennert at vanalboom.org> posited:

> I'm running into issues with disk shadowing in VMS between two SIMH instances.
> My setup is as follows:
> 
> - 2 SIMH VAX instances running VMS 7.3, fully licensed using hobbyist licenses:
>   WHISKY and COGNAC
> - Networking was setup using the "taptap" method as described in
>   http://www.retrocomputinggeek.com/retrowiki/SIMHNetworking
> - Both SIMH network tunX devices are on the same bridge (br0) on the same linux
>   host
> - DECnet and TCPIP configured and confirmed to be working on both machines:
>   WHISKY (1.1, x.y.z.240), COGNAC (1.2, x.y.z.241)
> - VAXcluster enabled between both instances: WHISKY has VOTES=1, COGNAC
>   VOTES=0, cluster EXPECTED_VOTES=1 with no quorum disk
> - ALLOCLASS for WHISKY is set to 240, COGNAC to 241
> 
> The Linux host itself is a Xen virtual machine; this is the main reason why I
> didn't go for a "dedicated" network interface per simh instance, as this
> doesn't play nice with the bridges on the Xen VM host.
> 
> At this point, I wanted to play with disk shadowing. Each of the SIMH VAXen has
> two disks; one for system (DUA0:), one 'spare' (DUA1:). Both are of type RA90
> (1GB). The idea was to shadow those two 'spare' disks.
> 
> Commands issued:
> 
> $ INIT/ERASE $240$DUA1: CLUSTERDATA
> $ INIT/ERASE $241$DUA1: CLUSTERDATA
> $ MOUNT/SYSTEM/CLUSTER DSA0: /SHADOW=($240$DUA1:,$241$DUA1:) CLUSTERDATA CLUSTERDATA
> 
> At this point, the shadow disk is being created and mounted on both nodes. No
> issues there. But the disk copy never actually kicks off:
> 
> $ SHOW DEV CLUSTERDATA
> 
> Device                  Device           Error    Volume         Free  Trans Mnt
>  Name                   Status           Count     Label        Blocks Count Cnt
>  DSA0:                   Mounted              0  CLUSTERDATA    2375847     1   2
>  $240$DUA1:    (WHISKY)  ShadowSetMember      0  (member of DSA0:)
>  $241$DUA1:    (COGNAC)  ShadowCopying        0  (copy trgt DSA0:   0% copied)
> 
> 
> Even after 12 hours, it's still at 0%; I think it's safe to say it's not going
> to go up. Has anyone ever attempted to do disk shadowing between two SIMH
> instances, and if yes, do you have any idea what might be causing this problem?

Patches for the OpenVMS VAX V7.3 version here (for shadowing and a general VMS update patch) would be one detail to look into.  (Patch access has been locked down for OpenVMS Hobbyists, unfortunately.)

The shadowing commands and sequence are (mostly) valid and (as we'd discussed offline) beyond building the shadowset one volume at a time and moving from MOUNT /CLUSTER to just MOUNT /SYSTEM on each host (and away from MOUNT /SYSTEM/CLUSTER), that all looks fine.

What's odd here is why the shadow copy doesn't get rolling, and that implies either something with VMS itself (hence patches) or with how simh is presenting the volumes or the network communications, or potentially with Xen.  (Something else that might not be "playing nice" with Xen here, too.)

And I'd try moving away from Xen and to simh directly on Linux directly on the hardware, as a test.

Stephen Hoffman
HoffmanLabs LLC