[Info-vax] Unpleasant Disk Shadowing Surprise

[abrsvc]

On Oct 12, 3:34 pm, JF Mezei <jfmezei.spam... at vaxination.ca> wrote:
> abrsvc wrote:
> > In a host based shadowing scenerio, a failure of a member WILL NOT
> > cause the entire DSA device to enter mount verification.
>
> But ! But !
>
> Rob Brooks confirmed that they do go into mount verify, a process
> necessary to determine if the remaining members of the shadowset are
> trustable (aka: valid data etc) before resuming service.
>
> The question should really be: what could cause mount verification to
> last 3 minutes.  (perhaps both disks were inaccessible during that time).

I will confirm that Rob was correct and I was wrong.  The code
explicitly DOES show that verification occurs.  In my experience with
shadowing, I have lost drives WITHOUT seeing the verification
directly, thus my statement.  I have NEVER seen a delay as described
in this thread initially unless all members of a set were lost.  I
just tried a few experiements here with a DS10 and shadowsets and
found the following:

1) The "failure" of a single member with little activity on the disk
resulted in an OPcom message and little visible delays.
2) The same experiment with an active disk, DID show a delay (although
less than 3 minutes).  I did not exactly time the delay, but it did
seem close to the mbr_timeout value referred to in other postings.

Dan