[Info-vax] Odd Show Device (DSSI disks in cluster)

[Michael Moroney]

JF Mezei <jfmezei.spamnot at vaxination.ca> writes:

>Volker Halle wrote:

>> if the MSCP device is served by 2 nodes, node 1 will know about that.
>> During MOUNT, it will try both pathes to the disk via node 2 and also
>> via node 3. If one path works, it will be able to mount the disk.

>Bot how does Node1 learn that Node 2, via whom it had mounted the drive,
>has lost connection to Disk A ? (so it can failover to node3 as MSCP
>server for that disk).

Node 1 DUDRIVER will issue a MSCP Packack (I forget the exact name of the
command at that level) to Node 2.  It will fail, and Node 1 will issue
another Packack to Node 3, where it will hopefully work if the device is
valid from there.  As Volker mentioned, there isn't any provision in MSCP
to say "this device has gone away" asynchronously, it's not until the
device is attempted to be used that an error status is returned.  There is
more complication at the PACKACK level trying to bring things online.

-Mike
(former DUDRIVER/MSCP Server maintainer)