Cluster status is not updated properly

[vexxhost]

Cluster status is not updated properly

[okozachenko1203]

• Cluster status is not changed from UPDATE_COMPLETE to UPDATE_IN_PROGRESS (the case which @guilhermesteinmuller mentioned above)
• Cluster status is not reflecting the failed result. Even though cluster creation failed because of resource lack, it keeps CREATE_IN_PROGRESS status.

[mnaser]

@okozachenko1203 so I think this issue should track the first part of your message, and the second part is covered in #64

[okozachenko1203]

Cluster status transition from COMPLETE to IN_PROGRESS is done by conductor's periodic service ClusterUpdateJob. mcapi driver only does update the status from IN_PROGRESS to COMPLETE.
This is more likely related to the race condition and the conductor was stuck to do the periodic service, and not related to mcapi.

[mnaser]

i think this issue still valid, if I start an upgrade o fthe cluster, it goes into COMPLETE right away because the reconciler did not start the upgrade yet.. we should somehow find a way to 'delay' the check or know that it will go into reconciliation

[okozachenko1203]

i think this issue still valid, if I start an upgrade o fthe cluster, it goes into COMPLETE right away because the reconciler did not start the upgrade yet.. we should somehow find a way to 'delay' the check or know that it will go into reconciliation

Add a delay could be the simple fix but not perfect i think
To know it will go into reconciliation, we need to share the k8s controller task queue or events? Not sure it could be possible.

Now in the mcapi driver's upgrade_cluster(), it just applies the desired states to the capi resources and ends. And the cluster_conductor's cluster_upgrade handler will hand over the task and update the cluster status to IN_PROGRESS. I think mcapi driver's upgrade_cluster() method can wait till the new reconciliation is started, so conductor will change the coe cluster status after the capi controller changed the capi cluster status.