Comment 5 for bug 1282842

Revision history for this message
Salvatore Orlando (salvatore-orlando) wrote :

The neutron error message reported at #2 actually also appears in gate logs:

http://logstash.openstack.org/#eyJzZWFyY2giOiJtZXNzYWdlOlwiUXVldWVQb29sIGxpbWl0IG9mIHNpemUgMTAgb3ZlcmZsb3cgMjAgcmVhY2hlZFwiIEFORCBOT1QgbWVzc2FnZTpcIlRyYWNlYmFjayAobW9zdCByZWNlbnRcIiIsImZpZWxkcyI6W10sIm9mZnNldCI6MCwidGltZWZyYW1lIjoiNjA0ODAwIiwiZ3JhcGhtb2RlIjoiY291bnQiLCJ0aW1lIjp7ImZyb20iOiIyMDE0LTAyLTIxVDA5OjI2OjAxKzAwOjAwIiwidG8iOiIyMDE0LTAyLTIxVDExOjI2OjAxKzAwOjAwIiwidXNlcl9pbnRlcnZhbCI6IjAifSwic3RhbXAiOjEzOTMxNjk1OTUzNzYsIm1vZGUiOiIiLCJhbmFseXplX2ZpZWxkIjoiIn0=

When this error appears, the job often (27% of cases) succeeds anyway. Also, the fact that it does not show up on the large ops job indicates the agents are exhausting the connection pool.

I suggest filing a different neutron bug for this specific issue, and add an E-R query as well, but this is not important.
The 500 errors returned from Nova might be because of this error affecting an API request, but also because of bug 1283522, which seems to actually be the most recurrent cause of error.