deepblue1968
Cadet
- Joined
- May 16, 2017
- Messages
- 5
My setup consists of two filers running FreeNAS-11.2-U7 with 20TB on the primary (lets call it A) and 40TB on the secondary (lets call it B), both in RAID-Z3.
On A I have configured one single POOL, with main DATASETS for APPLE, CIFS and UNIX.
Daily recursive periodic snapshots have been configured on all main datasets, snapshot lifetime set to 1 year. Including recursive sub-datasets, I end up with 20 individual snapshots per day. All snapshots get replicated every night to B via three recursive Replication Tasks.
This setup has been successfully running for 125 days, leading to 2.500 snapshots on both systems in full sync.
Today I had to perform some clean-up on A, since disk utilization had risen >80% and I decided to purge old snapshots from A.
Doing so for all year 2019 snapshots was unproblematic - all snapshots on B were left untouched, while some space was freed up on A.
Continuing in stages, I dared my luck and ultimately purged also ALL snapshots from year 2020 on A.
==> After that, I found that now all snapshots on B were gone as well! Literally the secondary filer is now empty - only the three main datasets still exit on B with 158 MiB used, instead of 17 TiB before.
Surely I must have misunderstood something in the way replication works!
Naively I had assumed that the option "Delete Stale Snapshots on Remote System" set to OFF would prevent the purging of the snapshots on B.
Is it correct to assume that at least ONE "generation" of snapshot of all datasets has to "survive" on the replication source A?
Otherwise, it appears that the replication logic will initiate a removal off all replicated snapshots on the target system...
I'm not in panic, since my data is still available on filer A, furthermore I have three independent full copies of the data on external disks and I have a third filer which was powered down at the time of the manipulation, holding an older set of replicated snapshots, so I'm pretty well covered.
But maybe somebody could point me to the part of documentation I have been missing, which explains above behaviour of replicated snapshots when their base is lost...
On A I have configured one single POOL, with main DATASETS for APPLE, CIFS and UNIX.
Daily recursive periodic snapshots have been configured on all main datasets, snapshot lifetime set to 1 year. Including recursive sub-datasets, I end up with 20 individual snapshots per day. All snapshots get replicated every night to B via three recursive Replication Tasks.
This setup has been successfully running for 125 days, leading to 2.500 snapshots on both systems in full sync.
Today I had to perform some clean-up on A, since disk utilization had risen >80% and I decided to purge old snapshots from A.
Doing so for all year 2019 snapshots was unproblematic - all snapshots on B were left untouched, while some space was freed up on A.
Continuing in stages, I dared my luck and ultimately purged also ALL snapshots from year 2020 on A.
==> After that, I found that now all snapshots on B were gone as well! Literally the secondary filer is now empty - only the three main datasets still exit on B with 158 MiB used, instead of 17 TiB before.
Surely I must have misunderstood something in the way replication works!
Naively I had assumed that the option "Delete Stale Snapshots on Remote System" set to OFF would prevent the purging of the snapshots on B.
Is it correct to assume that at least ONE "generation" of snapshot of all datasets has to "survive" on the replication source A?
Otherwise, it appears that the replication logic will initiate a removal off all replicated snapshots on the target system...
I'm not in panic, since my data is still available on filer A, furthermore I have three independent full copies of the data on external disks and I have a third filer which was powered down at the time of the manipulation, holding an older set of replicated snapshots, so I'm pretty well covered.
But maybe somebody could point me to the part of documentation I have been missing, which explains above behaviour of replicated snapshots when their base is lost...
