[Info-vax] Unpleasant Disk Shadowing Surprise

[Kenneth Fairfield]

On Wednesday, October 12, 2011 1:46:10 PM UTC-7, MG wrote:
> On 12-10-2011 21:36, Kenneth Fairfield wrote:
> > Having come from HSJ/CI-attached storage where no errors
> > like these get logged (on mount or boot), and having
> > similar experience as yours, I just took it to be
> > characteristic of FC-attached disks.  Whether it's the
> > device drivers, some sort of polling of the devices, or
> > whatever, they just seem to work that way.
> 
> The strange thing is, that the disks are served over two
> channels to two nodes and that only one of the controllers
> in one of the two served nodes records these errors.

If I were to take a guess, I'd say the error occurs on
the controller that has the active path to the disks.

In a cluster with storage configured on FC, it's common
for cluster "events", whether a node crashing or just
some less serious "blip", paths do the disks will change.
Those path switches show as an error on each disk that
switched paths.

There's lots more to all this.  My recollection is
that we forced paths to where we wanted them (balanced
across controllers and FC switches) before mounting the
disks at boot time.  In that case, I don't *think* an
error was logged, but I can't verify one way or the 
other (that cluster is gone).

   -Ken