How to safely clean up old dump files for freeing space?

[bzon]

I have run dragonfly with --save_schedule:*:10 (every 10 minutes). I noticed that it keeps a lot of snapshots by default. Is it safe to delete these files? If I will create a clean up script, should I just keep the latest snapshots (by timestamp)? Is

-rw-r--r-- 1 root 1001 323M Jul 18 15:10 dump-2023-07-18T15:10:17-0000.dfs
-rw-r--r-- 1 root 1001 322M Jul 18 15:10 dump-2023-07-18T15:10:17-0003.dfs
-rw-r--r-- 1 root 1001 322M Jul 18 15:10 dump-2023-07-18T15:10:17-0002.dfs
-rw-r--r-- 1 root 1001 322M Jul 18 15:10 dump-2023-07-18T15:10:17-0006.dfs
-rw-r--r-- 1 root 1001 322M Jul 18 15:10 dump-2023-07-18T15:10:17-0004.dfs
-rw-r--r-- 1 root 1001 322M Jul 18 15:10 dump-2023-07-18T15:10:17-0001.dfs
-rw-r--r-- 1 root 1001 322M Jul 18 15:10 dump-2023-07-18T15:10:17-0005.dfs
-rw-r--r-- 1 root 1001 4.0K Jul 18 15:10 dump-2023-07-18T15:10:17-summary.dfs
-rw-r--r-- 1 root 1001 336M Jul 18 16:10 dump-2023-07-18T16:10:04-0000.dfs
-rw-r--r-- 1 root 1001 335M Jul 18 16:10 dump-2023-07-18T16:10:04-0005.dfs
-rw-r--r-- 1 root 1001 335M Jul 18 16:10 dump-2023-07-18T16:10:04-0002.dfs
-rw-r--r-- 1 root 1001 335M Jul 18 16:10 dump-2023-07-18T16:10:04-0006.dfs
-rw-r--r-- 1 root 1001 335M Jul 18 16:10 dump-2023-07-18T16:10:04-0004.dfs
-rw-r--r-- 1 root 1001 335M Jul 18 16:10 dump-2023-07-18T16:10:04-0001.dfs
-rw-r--r-- 1 root 1001 4.0K Jul 18 16:10 dump-2023-07-18T16:10:04-summary.dfs
-rw-r--r-- 1 root 1001 335M Jul 18 16:10 dump-2023-07-18T16:10:04-0003.dfs

[romange]

hi @bzon . As explained by https://www.dragonflydb.io/docs/managing-dragonfly/backups our default mode is using dump-{timestamp} filenames. However, if you want you can override the flag and just use dump which will override the same snapshot again and again.

[romange]

what example @cristianrat ?
we have a partial example here: https://www.dragonflydb.io/docs/managing-dragonfly/backups#the-dbfilename-flag
though I agree it's not enough

[cristianrat]

I had assumed it's possible via k8s to just specify the name in the snapshot section, but from what I see so far (please correct me if I am wrong) I need to pass it as an arg?
this is how my terraform looks now

.......
spec:
  replicas: ${local.min_replicas}
  image: docker.dragonflydb.io/dragonflydb/dragonfly:${local.dragonfly_version}
  args: ["--dbfilename=snapshot_ftw"]
  authentication:
...

In the docs (which are reasonably good) it would be a good idea to make a note: pruning is not automatic and you have to deal with that yourself
I was caught out by this and my disk filled up (yes, you can say my APM is not proper, and it wasn't :) clearly )

[romange]

first of all - thanks for the suggestion! And its good idea to add this comment - you are not the first one who bit by that. You are more than welcome to send a PR, btw - our docs are located there:
https://github.com/dragonflydb/documentation - it can even account for hacktoberfest if you participate.

[romange]

you need to overwrite the dbfilename with a constant name like --dbfilename=snapshot to not have multiple snapshots. Unfortunately I do not know terraform to comment how it should be configured there.

[cristianrat]

already in prod, so that's the right way :) via terraform
if anybody wants the whole manifest, let me know, but should be clear enough
will raise a PR to the docs