Closed nictuku closed 8 years ago
My best guess is that the mgo connection is becoming stale so we need to refresh it if it becomes bad.
Working on this now
I didn't restart the database recently, so the bug didn't hit us recently.
But I went ahead and made a few partial fixes: 851b629c23f68f8533979a4e84da06eab5ea6e71 and d40f212a06319e6c40cec93719d05d0cd3802bba.
Those commits make the subscriber (and parts of the web server) to grab a new session when needed. Well, that's in theory. The mgo documentation isn't super clear about how and when will sessions be reconnected after the db is restarted.
Didn't process anything or 6 days. This happened before. A restart fixes it.
This is a stack dump: