As the bugfix is not fixing the race condition (and that's not trivial, due to how RT provides resources to the scheduler), it creates a side-effect issue.
Maybe we could try to improve the number of retries ? That said, 120 seconds for booting 3 nodes is still big IMHO.
http:// logstash. openstack. org/#eyJzZWFyY2 giOiJtZXNzYWdlO lwiTm92YSBoeXBl cnZpc29yLXN0YXR zIGRpZCBub3Qgcm VnaXN0ZXIgYXQgb GVhc3QgMyBub2Rl c1wiIiwiZmllbGR zIjpbXSwib2Zmc2 V0IjowLCJ0aW1lZ nJhbWUiOiI0MzIw MCIsImdyYXBobW9 kZSI6ImNvdW50Ii widGltZSI6eyJ1c 2VyX2ludGVydmFs IjowfSwic3RhbXA iOjE0MTc3ODQzOD I2NDZ9
As the bugfix is not fixing the race condition (and that's not trivial, due to how RT provides resources to the scheduler), it creates a side-effect issue.
Maybe we could try to improve the number of retries ? That said, 120 seconds for booting 3 nodes is still big IMHO.