Thank you for your knowledge. I have a question. Which pool is affected when the PG is down, and how can I show it? When a PG is down, is only one pool affected or are multiple pools affected? Vào Th 3, 26 thg 12, 2023 vào lúc 16:15 Janne Johansson < icepic.dz@xxxxxxxxx> đã viết: > Den tis 26 dec. 2023 kl 08:45 skrev Phong Tran Thanh < > tranphong079@xxxxxxxxx>: > > > > Hi community, > > > > I am running ceph with block rbd with 6 nodes, erasure code 4+2 with > > min_size of pool is 4. > > > > When three osd is down, and an PG is state down, some pools is can't > write > > data, suppose three osd can't start and pg stuck in down state, how i can > > delete or recreate pg to replace down pg or another way to allow pool to > > write/read data? > > > Depending on how the data is laid out in this pool, you might lose > more or less all data from it. > > RBD images get split into pieces of 2 or 4M sizes, so that those > pieces end up on different PGs, > which in turn makes them end up on different OSDs and this allows for > load balancing over the' > whole cluster, but also means that if you lose some PGs on a 40G RBD > image (made up of 10k > pieces), chances are very high that the lost PG did contain one or > more of those 10k pieces. > > So lost PGs would probably mean that every RBD image of decent sizes > will have holes in them, > and how this affects all the instances that mount the images will be > hard to tell. > If at all possible, try to use the offline OSD tools to try to get > this PG out of one of the bad OSDs. > > https://hawkvelt.id.au/post/2022-4-5-ceph-pg-export-import/ might > help, to see how to run > the export + import commands. > > If you can get it out, it can be injected (imported) into any other > running OSD and then replicas > will be recreated and moved to where they should be. > > If you have disks to spare, make sure to do full copies of the broken > OSDs and work in the copies > instead, to maximize the chances of restoring your data. > > If you are very sure that these three OSDs are never coming back, and > have marked the OSDs > as lost, then I guess > > ceph pg force_create_pg <pgid> > > would be the next step to have the cluster create empty PGs to replace > the lost ones, but I would > consider this only after trying all the possible options for repairing > at least one of the OSDs that held > the PGs that are missing. > > -- > May the most significant bit of your life be positive. > -- Trân trọng, ---------------------------------------------------------------------------- *Tran Thanh Phong* Email: tranphong079@xxxxxxxxx Skype: tranphong079 _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |