kubernetes bay needs to be fully functional before notifying the wait condition and transitioning to CREATE_COMPLETE

Bug #1551824 reported by Corey O'Brien
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Magnum
Fix Released
High
hongbin

Bug Description

Currently a kubernetes bay goes to CREATE_COMPLETE before the kubernetes API is fully active. The template only waits for all the start commands to be issued but doesn't wait for all the kubernetes setup work to complete. As a result we get kubernetes failures in the gate due to the tests trying to connect to kubernetes before it is actually ready.

Example stack trace from test: http://paste.openstack.org/show/488773/
Example gate failure from bay that went CREATE_COMPLETE before kubernetes was functional:
http://logs.openstack.org/05/275405/6/check/gate-functional-dsvm-magnum-k8s/4c1c488/
 * The bay went to CREATE_COMPLETE at 05:12:36 (from m-cond.log)
 * master node was still creating and running podmaster and proxy and other startup containers as of 5:16:34, 5:17:23, etc.

Changed in magnum:
assignee: nobody → Corey O'Brien (coreypobrien)
status: New → In Progress
Revision history for this message
Corey O'Brien (coreypobrien) wrote :

Magnum also needs to wait for minion nodes to be registered with the master

Revision history for this message
Ton Ngo (ton-i) wrote :

This also has implication when other tools interface with Magnum via the API. Gate is only one example.
We should ensure that when the bay status goes to CREATE_COMPLETE, the COE is fully functional.

Revision history for this message
Corey O'Brien (coreypobrien) wrote :

Agreed Ton. I want this issue to reflect the overall problem, not just the gate.

hongbin (hongbin034)
Changed in magnum:
importance: Undecided → High
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to magnum (master)

Reviewed: https://review.openstack.org/275405
Committed: https://git.openstack.org/cgit/openstack/magnum/commit/?id=08ac9a6cbac5b324747c178c4cab36296c16b567
Submitter: Jenkins
Branch: master

commit 08ac9a6cbac5b324747c178c4cab36296c16b567
Author: Corey O'Brien <email address hidden>
Date: Tue Feb 2 15:57:40 2016 -0500

    Remove minion dependency on master

    Make thhe minions work correctly without depending on the master.
    Removing this dependency decreases provisioning time by provisioning
    nodes in parallel.

    Change-Id: I025dbf54ef0cdafc4fe26acc33a9faed8cacd1ca
    Partial-Bug: 1536739
    Partial-Bug: 1551824

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to magnum (master)

Fix proposed to branch: master
Review: https://review.openstack.org/301954

Changed in magnum:
assignee: Corey O'Brien (coreypobrien) → hongbin (hongbin034)
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to magnum (master)

Reviewed: https://review.openstack.org/301954
Committed: https://git.openstack.org/cgit/openstack/magnum/commit/?id=8972b0c1c4a45cf8fd766faeb0e3db0bcc51354c
Submitter: Jenkins
Branch: master

commit 8972b0c1c4a45cf8fd766faeb0e3db0bcc51354c
Author: Hongbin Lu <email address hidden>
Date: Tue Apr 5 18:18:06 2016 -0400

    Notify Heat only if kube-apiserver is running

    Kubernetes bay needs to be fully functional before notifying Heat
    to bring the stack to CREATE_COMPLETE. This patch added a systemd
    unit that keeps pinging the kube-apiserver and notify Heat once
    it goes through.

    Change-Id: Ib4fb22f3040d76d488c6cac6873d64b4d973ebb1
    Closes-Bug: #1551824

Changed in magnum:
status: In Progress → Fix Released
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix included in openstack/magnum 3.0.0

This issue was fixed in the openstack/magnum 3.0.0 release.

Revision history for this message
Spyros Trigazis (strigazi) wrote :

I think we need to open this bug again, heat should be notified when scheduler and controller-manager are ready.

Thoughs?

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.