Runpod•5d ago

So whats the deal with all the issues and the charging for failed docker fetches?

This has been going for over 2 days now, where is the statement? I'm kinda losing all the trust i had built up with runpod. The biggest issue are the charges that you have to notice and stop manually on failed docker fetches, thats an absolute no-go. Issues: - Workers CHARGING me and IGNORING the execution timeout when they somehow manage to take a job, but are stuck/fail to fetch the custom dockertemplate. (Stuff like this should be included in the execution timeout. My worker was running past 10 minutes, even tho i set the timeout to 200s. If i would've not noticed this, it probably would still be running!) - Workers getting stuck initializing (image pull: docker.io/...: pending) - Workers constantly switching, sometimes before they are even initialized. The charging and not respecting timeouts is the biggest issue in my eyes, you can't just allow something like this to happen.

7 Replies

Accai•5d ago

+1, i don't see why we would be charged for bugs originating from runpod on workers that usually work. Last time, a worker ran for 6 minutes (compared to an average of 20 seconds) because there was a DockerHub authentication bug (later reported as a bug). (my timeout is 120s btw)