I have been investigating neutron's connection as well, since I had an exceptionally high failure rate with neutron jobs.
In neutron logs there is nothing which might point to a failure leading to an instance not being spawn. More precisely, I have been looking for neutron API calls not returning or returning with error codes which might have caused the thread processing network info asynchronously to never end. However, my research was inconclusive.
Failure vs build names [1] shows a neutron job joint 1st in the highest ranked failing jobs. I am not sure however whether these numbers can confirm a skew in the failure rate towards neutron.
I have been investigating neutron's connection as well, since I had an exceptionally high failure rate with neutron jobs.
In neutron logs there is nothing which might point to a failure leading to an instance not being spawn. More precisely, I have been looking for neutron API calls not returning or returning with error codes which might have caused the thread processing network info asynchronously to never end. However, my research was inconclusive.
Failure vs build names [1] shows a neutron job joint 1st in the highest ranked failing jobs. I am not sure however whether these numbers can confirm a skew in the failure rate towards neutron.
[1] http:// logstash. openstack. org/#eyJmaWVsZH MiOltdLCJzZWFyY 2giOiJtZXNzYWdl OlwiRGV0YWlsczo gVGltZWQgb3V0IH dhaXRpbmcgZm9yI HRoaW5nXCIgQU5E IG1lc3NhZ2U6XCJ 0byBiZWNvbWVcIi BBTkQgIChtZXNzY WdlOlwiQUNUSVZF XCIgT1IgbWVzc2F nZTpcImluLXVzZV wiIE9SIG1lc3NhZ 2U6XCJhdmFpbGFi bGVcIilcbiIsInR pbWVmcmFtZSI6Ij YwNDgwMCIsImdyY XBobW9kZSI6ImNv dW50Iiwib2Zmc2V 0IjowLCJ0aW1lIj p7InVzZXJfaW50Z XJ2YWwiOjB9LCJt b2RlIjoic2NvcmU iLCJhbmFseXplX2 ZpZWxkIjoiYnVpb GRfbmFtZSJ9