[GLLUG] SAN events kill Linux filesystems

[Henrik Morsing]

Good morning,

Don't know if anyone has come across this but every time a SAN node is powered down on our SAN, even if un-used, we have a handful of Linux LPARs (across multiple frames) with filesystems going into read-only.

We had migrated away from our IBM V9000 and many months later powered it off resulting in the first occurrence of this issue. Later on, we patched our new EMC 9200T, which (I'm not a storage person) has two... linked systems?... with two nodes each, so four node in total. Even when the first node rebooted, we had a handful of Linux systems going into read-only (Other OSes un-affected).

I raised it with SUSE who kept blaming VIO NPIV mappings and settings without having anything concrete or any relevant suggestion. They have now given up.

My gut feeling is a driver or multipath software freaking out for no reason. When I checked the code for the ibmvfc module, it doesn't appear to have been updated since it was written in 2008, so a bit odd if no-one had noticed this by now. So, multipath maybe? Or something completely different?

Any thoughts?

- 
Regards,
Henrik Morsing