>From at least Nautilus and above auto repair is safe. If you have filestore OSDs in your cluster, auto repair will only run if auto_repair_supported() is true for your backend. This is true in the case of EC (regardless of filestore/bluestore). Or replicated with has_builtin_csum() = True, which would be Bluestore only. On Tue, May 10, 2022 at 12:34 AM Janne Johansson <icepic.dz@xxxxxxxxx> wrote: > Den mån 9 maj 2022 kl 21:04 skrev Vladimir Brik > <vladimir.brik@xxxxxxxxxxxxxxxx>: > > Hello > > Does osd_scrub_auto_repair default to false because it's > > dangerous? I assume `ceph pg repair` is also dangerous then? > > > > In what kinds of situations do they cause problems? > > With filestore, there were less (or no?) checksums, so the cluster > might not always have an idea which replica was correct if they > differ, so auto-repair on those would in theory copy the error around > in the worst case. For bluestore there are checksums, so the chances > of your cluster detecting an error somewhere, while one or more of the > replicas have another undetectable error that still gives a correct > checksum so that this error gets copied around feels somewhat less > likely. > > > -- > May the most significant bit of your life be positive. > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx > To unsubscribe send an email to ceph-users-leave@xxxxxxx > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |