dvc9
Explorer
- Joined
- May 2, 2012
- Messages
- 72
Hey!
Think I found a possible bug in the Replication Tasks, ill try to fill out the details here as good as I can,
and hopefully that can be a full bug report I can add and contribute to Jira later on!
The setup
A - local Production
Running Periodic Snapshot Tasks, on selected datasets,
We are keeping snapshot for 3 days on this server,
and the frequency is one per day.
Running Replication Tasks, To B - local Backup on all datasets.
B - local Backup
Getting datasets, from A - local Production.
Keeping the clone and all snapshots as long as possible, on this server.
Have open up the possibility to browse the .zfs/snapshots so I can quickly go back in time and pick tings up.
Replication Tasks has this settings :
Recursively Replicate Child Dataset Snapshots = checked off
Delete Stale Snapshots on Remote System = checked off
Replication Stream Compression = off
Encryption Cipher = disabled
Dedicated User Enabled - remote key added.
The bug or not.
The B - local Backup server swapped location the other day, and the the Replication Tasks where off with 5 days.
B - local Backup, where turned on, and Replication Tasks on A - local Production started running.
The replication job did not pickup from the latest snapshot stored on B - local Backup , and took it from there,
the Replication Tasks jobs deleted all snapshots, and it feelt like it started "clean", like they never have been in sync before.
------------------------------------------------------------------------------------
So I do have a working theory, and maybe this is something to do with ZFS.
This will be a bit hard to explain, but ill try....
When Replication Tasks sends a incremental backup from A - local Production to B - local Backup
Then both servers need to have the same snapshot starting point
So if B has the data " FULL + Snapshot -- A1, A2, A3, A4, A5 "
and A has data " FULL + Snapshot -- A5, A6, A7 "
Since both servers has Snapshot A5 then it works
So if B has the data " FULL + Snapshot -- A1, A2, A3, A4, A5 "
and A have gone out of sync, with B, and has data " FULL + Snapshot -- A8, A9, A10 "
Since B is missing A6, and A7 then A is out of sync, and deletes all datasets on B, and starts a clean sync.
------------------------------------------------------------------------------------
Is this a bug, or a feature request ?
Think I found a possible bug in the Replication Tasks, ill try to fill out the details here as good as I can,
and hopefully that can be a full bug report I can add and contribute to Jira later on!
The setup
A - local Production
Running Periodic Snapshot Tasks, on selected datasets,
We are keeping snapshot for 3 days on this server,
and the frequency is one per day.
Running Replication Tasks, To B - local Backup on all datasets.
B - local Backup
Getting datasets, from A - local Production.
Keeping the clone and all snapshots as long as possible, on this server.
Have open up the possibility to browse the .zfs/snapshots so I can quickly go back in time and pick tings up.
Replication Tasks has this settings :
Recursively Replicate Child Dataset Snapshots = checked off
Delete Stale Snapshots on Remote System = checked off
Replication Stream Compression = off
Encryption Cipher = disabled
Dedicated User Enabled - remote key added.
The bug or not.
The B - local Backup server swapped location the other day, and the the Replication Tasks where off with 5 days.
B - local Backup, where turned on, and Replication Tasks on A - local Production started running.
The replication job did not pickup from the latest snapshot stored on B - local Backup , and took it from there,
the Replication Tasks jobs deleted all snapshots, and it feelt like it started "clean", like they never have been in sync before.
------------------------------------------------------------------------------------
So I do have a working theory, and maybe this is something to do with ZFS.
This will be a bit hard to explain, but ill try....
When Replication Tasks sends a incremental backup from A - local Production to B - local Backup
Then both servers need to have the same snapshot starting point
So if B has the data " FULL + Snapshot -- A1, A2, A3, A4, A5 "
and A has data " FULL + Snapshot -- A5, A6, A7 "
Since both servers has Snapshot A5 then it works
So if B has the data " FULL + Snapshot -- A1, A2, A3, A4, A5 "
and A have gone out of sync, with B, and has data " FULL + Snapshot -- A8, A9, A10 "
Since B is missing A6, and A7 then A is out of sync, and deletes all datasets on B, and starts a clean sync.
------------------------------------------------------------------------------------
Is this a bug, or a feature request ?
Last edited:
