Bug #21751
OSD crashes during recovery on arm64 due to assert in Throttle::put
0%
Description
Hey, been playing with building a cluster out of pine64 boards and have been getting constant OSD crashes after pools reach a size of ~130gb, due to an assert in Throttle::put. I'm using deb packages I built myself using Scaleway's arm64 platform to get around the extant crc32c bug in the released arm64 debs (see http://tracker.ceph.com/issues/19705).
I can't attach the crash dump because of a size limit in the tracker, so please accept a dropbox link instead, containing the logged info at the time of the crash and the output of objdump as requested in the log: https://www.dropbox.com/sh/f2zetrcg86t3gia/AADM4r29yM80Um5YYF5blUHqa?dl=0
Please let me know what else would be helpful to debug this.
History
#1 Updated by Peter Woodman over 6 years ago
I realized I forgot to mention- this is a crash loop. Once this crash starts happening, it doesn't stop, and no progress can (seemingly) be made towards a healthy cluster. After being restarted by systemd and finishing OSD boot, it crashes again within ~3 seconds, so it's, well, fairly crippling.
#2 Updated by Sage Weil almost 3 years ago
- Status changed from New to Closed