Closed j0nes2k closed 10 years ago
Check your Emperor logs, if they reports the vassal is not responding to heartbeats, you have hit a "timing bug" with the heartbeat subsystem. Just upgrade to 2.0.3 and the issue should disappear. If you cannot upgrade, remove heartbeat from the vassal
Upgrading to uWSGI 2.0.3 seems to have done the trick - at least in the first test. I will report back tomorrow and close this issue if everything is still working fine.
I can confirm that this was the error. Thank you for your help, closing this issue.
I have uwsgi v2.0 running on Ubuntu 13.04. uwsgi runs in Emperor mode with two apps deployed. The apps share a large amount of codebase, one is an API app, the other one is a Flask- and Jinja2 based webapp. The API app has been running fine for the past few months without any problems.
Now we are in the process of deploying the webapp. This app behaves strange - in the logs I can periodically see vassals complaining about disconnecting masters, going into some kind of "respawn loop". The app itself does not report any errors and after its restart (which takes 4-5 seconds) it is responding fine again. This happens with or without requests hitting the webapp. Without any request, uwsgi respawns about every 2 minutes. With a simple test (hitting the same page in 2 second intervals), the app is running about 10-15 minutes without respawning.
There should be no exception thrown from the app, and in the logfiles and stats daemons everything is looking fine.
What is the cause for this behaviour? Any idea on how to debug this further?
Here is the emperor config:
...and here is the webapp config:
Excerpt from the logfile /var/log/uwsgi/frontend.log: