I have watchtower to keep my 50 production docker containers up to date. So far I have never had to resort to an old image since the new one was bugged but I know of the risks and dont have the --cleanup flag activated so I could easily spin up the old image if necessary. I also keep daily (mirrored) backups so I should be ok in the case of failure.
But I keep running into space issues due to multiple GBs of old images that I have to manually remove to not fill up the rather small ssd.
Does anyone have an idea how to automatically remove all but the newest unused docker image so I can quickly revert to it in case of failure but once there is another unused one of this container it goes away?
It’s probably no big deal but searching for it together with watchtower didnt yield anything so far.
Thanks in advance and have a good one.
You must log in or register to comment.
Not to my knowledge, but you can set up a cron job to prune containers older than X days.
Yeah, I found the docker system prune filter option in the docs as well. Very unprecise sadly but thanks for mentioning it. :) I’m thinking of something like “when updating, first copy the version number of the old image and exclude it from prune”. Probably going to have to write something myself again. Feel free to spitball with me.
So my thought with the time based pruning is that you can keep a backup that’s X days old.
Let’s say you keep 2 weeks. If there have been no problems with an image after 2 weeks of an update, you’re probably good to go. If you have an issue during those 2 weeks, you can return to and image within those 2 weeks. If you’ve had no problems after 2 weeks, it’s probably stable.
Adjust 2 weeks to whatever you’re comfortable with.
This would absolutely make sense, were it not for the fact that the old image can be 3 weeks old when the new one comes out. Feel free to correct me but I think a time based option on age is not sufficient.
Correct, that would not work for that case.
But thanks for suggesting and spitballing with me.
Not to sound flippant, but it seems like a solution looking for a problem. I use the --cleanup flag, and if there’s an issue, rolling back is as simple as changing dockerimage:latest to dockerimage:version that worked.
Unless I’m missing something.
I agree. If OP is scared that the image creator is clearing the old images, then OP should just mirror the registry or just have a system backup.
Pretty much this. I don’t even bother with watchtower anymore. I just run this script from cron pointed at the directory I keep my directories of active docker containers and their compose files:
#/bin/sh for d in /home/USERNAME/stacks/*/ do (cd “$d” && docker compose pull && docker compose up -d --force-recreate) done; for e in /home/USERNAME/dockge/ do (cd “$e” && docker compose pull && docker compose up -d --force-recreate) done;
docker image prune -a -f
Not exactly what you’re asking but you can push specific images to a private repo to keep specific versions… Then you can just use the cleanup tag or prune to clear them off the system and if you want to pull them again it won’t need to download it from the internet
Hmmmm… thats pretty cool. Thanks for the suggestion.
