Hi, we have an object storage cluster with a lot of backup data. The data is currently in an extra EC pool (jerasure, k=5,m=3), with a custom s3 storage_class called `backup-center-ec`. Those are dedicated buckets for backup jobs and make around half the data of a 2PB cluster. We want to create a new EC pool with the ISA plugin and maybe even adjust the k/m values. We hope to get a little bit more performance out of it. Because we have this EC pool only in one zonegroup, the backup script checks if the `backup-center-ec` class is available (or we even hardcoded it), so we came up with the following idea: 1. create a new pool with correct EC profile 2. rename `backup-center-ec` to `backup-center-ec-legacy-01` 3. recreate the `backup-center-ec` storage class with the new pool This should make it possible that new backups will go into the new pool, and the old backups will slowly rotate out of the old pool. IIRC reading an object does not requrie to know in which storage_class the object is stored. If this would work, I am already very happy. But maybe the migration could be spead up or forceful move data to the new pool? I thought I can maybe set the explicit placement for a bucket to the new pool and do a radosgw-admin bucket rewrite. Or are there other ways? Would that work? Does someone have experience with it? Cheers Boris _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |