Bug #14255
openqa: we are filling smithi disks with ffsb workloads
0%
Description
http://pulpito.ceph.com/gregf-2016-01-04_11:49:51-fs-master---basic-smithi/13310/
http://pulpito.ceph.com/gregf-2016-01-04_11:49:51-fs-master---basic-smithi/13291/
So, apparently we fill up even more space than anticipated. It's not on every run so we might just be able to turn it down a little bit...
Updated by Greg Farnum about 8 years ago
- Priority changed from High to Urgent
I haven't looked yet but I suspect we're just running for a set period of time and the smithis are so much faster than the HDD-based nodes that they manage to entirely fill the drives. So the cheap solution is to just turn that timer down everywhere, or maybe we can configure differently based on machine type...
Updated by Zheng Yan about 8 years ago
ffsb (config is at qa/workunits/suites/random_write.32.ffsb) only use about 13G space, no matter how long it runs. I checked the log, OSD is on top of btrfs. I think it's caused by btrfs' long-standing premature ENOSPC issue.
Updated by Greg Farnum about 8 years ago
- Priority changed from High to Low
This is still happening (http://pulpito.ceph.com/teuthology-2016-03-07_18:04:01-fs-master---basic-smithi/45723/, http://pulpito.ceph.com/sage-2016-02-25_18:19:03-fs-master---basic-smithi/27939/ et al) but I'm not sure there's much we can do about it excpet wait. :/