Hi Chris, this patch fixes potential data corruption indeed. But IMO its probabilty is pretty low and it rather tend to occur when performing bulk RocksDB/BlueFS writes during e.g. omap naming scheme update. Which in turn rather occurs between major point release upgrade. At least that's the scenario when I faced that. Given the above and pretty infrequent complains about such data corruptions I would it's fine to perform your upgrade on a regular schedule. IMO absolutely no rush. Thanks, Igor On 3/19/2021 12:53 PM, Chris Palmer wrote:
When looking over the changelog for 15.2.10 I noticed some bluefs changes. One in particular caught my eye, and it was called out as a notable change:
os/bluestore: fix huge reads/writes at BlueFS (pr#39701 <https://github.com/ceph/ceph/pull/39701>, Jianpeng Ma, Igor Fedotov)
It wasn't obvious to me how important that one was for data integrity. We'll roll 15.2.10 onto our clusters shortly, but would appreciate an opinion on whether any delay in installing this would increase the risk of anything nasty (e.g. silent data corruption). If that was the case we will push the release through rather more urgently!
Thanks, Chris
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io