At about 1:17am this morning, I started seeing "XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)" and it just kept spewing that out until I rebooted at 9:30am. The machine is a dedicated file server with nfs and samba exports. The time would probably correspond with when the other servers are backing up to this one. The servers that import file systems from this one via nfs just sat there stuck until the reboot.
Does anyone have any suggestions to resolve it?
Possibly relevant information gleaned from a bit of googling
- I'm using the default xfs parameters i.e. not messing with the block sizes. http://tracker.ceph.com/issues/6301 talks about an already applied patch to centos7 but seems to boil down to messing with the block sizes.
- single xeon E5-2609v2, 32gb ram, 7x4g sas raid 5 on lsi controller with hot spare (getinfo for disks below)
- not running ceph
- drbd is replicating and status was ok