Hi all, I am trying to collect information about index resharding of replicated buckets in the multisite rgw environment.
Both my (master and secondary) clusters are v13.2.6 Mimic After many tests I found a working solution (like below): 1) disable bucket synchronization on master zone [root@master]# radosgw-admin bucket sync disable --bucket=bucket1 2) wait for the metadata to propagate on each zone [root@master]# radosgw-admin bucket sync status --bucket=bucket1 ... Sync is disabled for bucket bucket1 [root@secondary]# radosgw-admin bucket sync status --bucket=bucket1 ... Sync is disabled for bucket bucket1 3) purge bucket on (each) secondary zone [root@secondary]# radosgw-admin bucket rm --purge-objects -- bucket=bucket1 4) reshard bucket on master zone [root@master]# radosgw-admin bucket reshard --bucket=bucket1 --num- shards=8 5) wait for the metadata to propagate on each zone (the same bucket instance id on each site) 6) enable bucket synchronization on master zone [root@master]# radosgw-admin bucket sync enable --bucket=bucket1 I based on the following articles: https://access.redhat.com/documentation/en-us/red_hat_ceph_storage/3/html-single/object_gateway_guide_for_red_hat_enterprise_linux/index#manually-resharding-buckets-with-multisite-rgw https://cloud.tencent.com/developer/article/1423086 The biggest disadvantages are: - Possible long time of bucket removal on the secondary zones (time between points 3 and 4) - At the end, all the files of bucket are synchronized from scratch Is there any (refreshed) recommended procedure of "bucket index resharding" in multisite environment? Is it possible to reshard bucket index without removing bucket on secondary zone? Best regards, PO _______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
