[Info-vax] Shadowing & Reboot

[Michael Moroney]

gartmann at nonsense.immunbio.mpg.de (Christoph Gartmann) writes:

>I created a shadow set with two disks, each disk connected to a different node.
>So far so good. Then I rebooted node 1. When node 1 was up again, I had to
>manually add its disk back into the shadow set. So, should I put this mount
>command into SYSTARTUP_VMS.COM? Next, the remounted disk went into a full
>copy state, although there were no changes made to the surviving disk on
>node 2. Is this normal?

1) If the shudown procedure dismounted the shadowset member as part of the
   normal shutdown, this is normal and expected.

2) If the system shut down without removing the member, the other 
   system(s) will wait SHADOW_MBR_TIMEOUT seconds before they throw the
   member out as unavailable.  In that case it has to be remounted into
   the shadowset.

In the second case, increasing the SYSGEN parameter SHADOW_MBR_TIMEOUT
on all systems to a value long enough to survive the other system being 
shut down (including writing a dump if a crash) and rebooting to a point
where the MSCP server makes the drive available will prevent this.
Note that file accesses to the shadowset may (will for writes) hang until 
either the drive becomes available or SHADOW_MBR_TIMEOUT seconds passes,
so there is a tradeoff between making it too short or too long.