Due to repeated intermittent hardware failures of labstore1002 disk system, we need to be able to perform tests without putting the actual live shelves at risk; so we need at least 2-3 shelves connected to the H800 controller once we are done with the raid10 migration.
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | yuvipanda | T105720 Labs team reliability goal for Q1 2015/16 | |||
| Resolved | coren | T106479 Ensure that labstore machine is 'known good' hardware | |||
| Resolved | • chasemp | T98183 labstore1002 issues while trying to reboot | |||
| Resolved | • chasemp | T101741 Locate and assign some MD1200 shelves for proper testing of labstore1002 | |||
| Resolved | coren | T96063 Migrate Labs NFS storage from RAID6 to RAID10 | |||
| Resolved | coren | T101011 Rsync live labstore filesystem to local eqiad copy | |||
| Resolved | mark | T101010 Make a block-level copy of the codfw mirror of labstore1001 to eqiad |
Event Timeline
Comment Actions
We do not have spare md1200 shelves lying around. I have one that is not used that is waiting to be connected to holmium. We may be able to temporarily use that. Is this still needed?
Comment Actions
What's going on with this? Title makes it seems like no proper testing has been done.
Comment Actions
Correct; this does need to be done - ideally before things explode and we have to move to it. @Cmjohnson, is that shelf intended for holmium still around?
Comment Actions
Did this happen? Does this still need to happen with the new labstore stuff that @chasemp / @madhuvishy have been working on?
Comment Actions
This task was murky when it was created to begin with. Closing it makes sense at this point.