Long running zencoder_fetchers die over time

[bcmreed]

We had set up fetchers in our QA environment, using the -l option, to constantly pull notifications.

However, overtime the process would crash. Sometimes several days later.

We used several instances of zencoder_fetcher, as we support multiple API keys. There did not seem to be a pattern of api key usage to the amount of jobs that were created. As in, I did not see a correlation of busy accounts to crashing the fetcher.

I don't have a stack trace as these were run as background jobs > /dev/null and it takes a long time to reproduce.

I will try to leave one running to get a trace, but I wanted to at least raise the issue.

I've moved away from the -l option and now use cron to trigger the script.

[chriswarren]

Thanks for the info. I don't think this'll be something we look in to fixing though, as Fetcher's really only been designed as a dev tool for doing short term integration testing, not working in QA or production environments.

If the stack trace turns up anything interesting that something can be done about quickly we'll see what we can do (or pull requests are welcome) but otherwise I doubt that we'll be able to make any significant investigation in to addressing this.