An update on the 2nd issue where VMs fail to launch: Gerry has confirmed the issue is due to using a nova docker image from OpenStack master. Don Penney is updating the docker image builds to use the OpenStack Stein branches. After re-testing confirms these docker images are sane, he will work to switch the CENGN builds over to the Stein branches.
Frank
From: Miller, Frank Sent: Wednesday, April 03, 2019 10:48 AM To: 'starlingx-discuss@lists.starlingx.io' starlingx-discuss@lists.starlingx.io Subject: Sanity updates
Folks:
I took an action on the containers community call to send out an update on the current sanity issues.
1. AIO-SX: This configuration should now be ready for use.
* Bart Wensley solved LP 1820928https://bugs.launchpad.net/starlingx/+bug/1820928 which turned out to be a bug in kubelet where it was hitting a limit of 250 http2 streams in a single connection.
2. Other multi-server configs: An intermittent issue still exists when launching VMs resulting in the VMs failing to be scheduled. Tracked under LPs 1821841https://bugs.launchpad.net/starlingx/+bug/1821841 & 1822116https://bugs.launchpad.net/starlingx/+bug/1822116
* Gerry Kopec continues to investigate intermittent issues with the nova-placement pod. When issue occurs VMs cannot be launched.
* Issue is with nova-compute unable to get requests processed to one of the nova-placement pods running on each controller.
* Current theory is this is related to our docker images using OpenStack master and a recent nova commit in the nova placement area is impacting the placement pod. Gerry expects to prove or disprove the theory later today.
Frank