Do you mean I/O stopped on your VMs? Sent from mobile

Op 15 aug. 2020 om 17:48 heeft Matt Dunavant <mdunavant@convokesystems.com> het volgende geschreven:

Hi all,

We just completed maintenance on an OSD node and we ran into an issue where all data seemed to stop flowing while the node was down. We couldn't connect to any of our VMs during that time. I was under the impression that by setting the 'noout' flag, you would not get the rebalance of the data but you would update the pointers to use the 2nd and 3rd copies of the data. Is that not correct, and what is the proper workflow for taking down an OSD node for maintenance?

Thanks, _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io

Hello all, Thanks for the help. I believe we traced this down to be an issue with the crush rules. It seems somehow osd_crush_chooseleaf_type = 0 got placed into our configuration. This caused ceph osd crush rule dump to include this line ' "op": "choose_firstn",' instead of 'chooseleaf_firstn'. Which I believe would tell ceph to make its failure domain an OSD instead of a node.