Support keepLast retention in k8up

[seanhamlin]

k8up supports a retention key keepLast https://k8up.io/k8up/1.2/how-tos/schedules.html

If would be nice if there was a way to pass these args into the KBD to then pass into the k8up schedule object.

Related to uselagoon/remote-controller#92

[Schnitzel]

can you elaborate the idea behind this? also we could make this the default? then we don't have to add additional arguments to KBD, helmcharts, etc/

[seanhamlin]

Yeah sure, so the idea is we have a customer that is looking to create 6 hourly backups. The current retention keys are a little confusing.

We currently can use

keepHourly: 72

Where I would prefer to have

keepLast: 12

To keep the last 3 days worth of 6 hourly backups.

also we could make this the default?

Potentially, likely it would still need to be combo'ed with keepMonthly etc if longer term backups are needed.

[Schnitzel]

mhh I'm not sure I understand, so if you say create 6 hourly backups you mean a backup every 6 hours? or do backups every hour and want to keep the last 6 hourly ones?

so assuming you do a backup every 6 hours and you want to keep them the last 3 days (4 backups per day), then keepHourly: 72 would should work, as the documentation (https://restic.readthedocs.io/en/latest/060_forget.html) says:

--keep-hourly n for the last n hours in which a snapshot was made, keep only the last snapshot for each hour.

we don't have 72 backups but based on how I read this this would keep hourly backups it could find in the last 72h, which would be all backups that have been created.

So I'm not sure why we need to add the keepLast feature as it currently can be implemented with what we have?

[seanhamlin]

you mean a backup every 6 hours?

Yes I do.

And yes this is achievable with the current functionality, but it did take a lot of reading of restic and k8up docs to discern this. If you feel this is not confusing, then feel free to close this issue.

[Schnitzel]

I'm more concerned adding a new functionality to a system that already works, I would rather fix actual bugs or implement features that don't work yet :)

But I let @tobybellwood decide on this