Houston UI Task Scheduler Replication task has a number of issues.
The target is a TrueNAS Scale system.
The pool dataset has recursive snapshots taken for roll back purposes
The RockyLinux Houston replication task expects the pool to be empty. Once there is a snapshot, it is no longer empty.
So tell TrueNAS side snapshot task to pass over the target dataset.
Pool is still dirty and task cannot be configured.
Create an empty dataset with no snapshots in the passed-over branch of the target pool. By the time the Houston UI side replication task form is completed, there is schmutz in the target dataset and task won’t add.
Messages are ambiguous about pool vs dataset and what is in the dataset, empty snapshots or actual data. Care regarding indefinite antecedents is needed.
Taking of empty snapshots is recommended for mishap or malice recovery.
I’m not sure how to get out of this one. Any ideas? Is a ZFS recommended snapshotting practice needed? How would it be incorporated into UI’s? How would it be communicated to admins?
There are recommended practices for taking and retention of weekly, monthly, and yearly snapshots and retention of same. Neither Houston or TrueNAS will create snapshots per this design pattern. The admin must do it. And this pattern is intent-free.
Brave’s buddy Leo has a pretty good explaination of what is going wrong and the mitigation.
Leo appears to be describing a somewhat different replication dialog than that I’m seeing,
Chat with Leo regarding replication between Rocky Linux and TrueNAS Scale
Leo suggests pushing from source is OK and maybe preferred as it gives a central view of all replications of the dataset.
Leo suggested -F to force the first replication. This would clean the slate on the target. All target dataset snapshots are discarded and all target data is replaced.
Leo recommends that the target not have a periodic snapshot task enabled as it will potentially loose the replication snapshots in the haystack. Each replication creates a target pool snapshot.
TrueNas replication task snapshots are named something like “auto-date”
Leo says the replication dialog lets you pick the snapshot to replicate. I didn’t see a UI element for that purpose.
Recommendations to Houston UI team:
- Each form should have a revision number so we can use the correct instructions. And so Houston UI boffins can identify the version being griped.
- I’d like to see a sandboxed Houston UI account used by Houston UI cron task daemons – nologin
- Test SSH connection notifications be saved on a stack or list for review. They are not being delivered to MacOS notification interface on the menubar. Portainer is a good model.
- Replication task makes the source and destination snapshots. This sort of slips by.
- Is the -F replication option needed for the first go? Adding a new dataset to the target pool on TrueNAS gives the dataset not empty exception. Even with the snapshot task disabled.
- Be careful to tease apart exceptions. Handle them call by call so users can determine the proper corrective action. And maintainers can do proper triage.
Hey Dismal,
I did just respond elsewhere so I am here again lol. I will pass this onto the team!