Bug 1956921

Summary: Deployments take a very long time and replication controller state gets out of sync
Product: OpenShift Container Platform Reporter: Luke Stanton <lstanton>
Component: kube-controller-managerAssignee: Maciej Szulik <maszulik>
Status: CLOSED NOTABUG QA Contact: zhou ying <yinzhou>
Severity: urgent Docs Contact:
Priority: unspecified    
Version: 3.11.0CC: aos-bugs, mfojtik, sttts
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2021-05-05 14:29:03 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Luke Stanton 2021-05-04 16:34:27 UTC 
Description of problem:

When creating a new deployment or scaling/refreshing an existing one, deployments take a very long time and may timeout. Symptoms also include pods being stuck in a 'Pending' or 'Terminating' state for a long time and replication controllers not reflecting the current/desired number of pods correctly, for example:

~~~~~~~~~~~~~
$ oc get pods
NAME             READY   STATUS        RESTARTS   AGE
httpd-5-j24x4    0/1     Terminating   0          15m
httpd-8-deploy   0/1     Error         0          12m
httpd-8-rlpzs    1/1     Running       0          12m
httpd-8-sgdcs    1/1     Running       0          5m
$ oc get rc
NAME      DESIRED   CURRENT   READY   AGE
httpd-7   1         1         1       21m
httpd-8   0         0         0       12m
$ oc get dc
NAME    REVISION   DESIRED   CURRENT   TRIGGERED BY
httpd   8          1         0         config,image(httpd:2.4)
~~~~~~~~~~~~~

This appears to be accompanied with slow etcd performance and consistent api errors related to pod listings.

Version-Release number of selected component (if applicable):

OCP 3.11.404


How reproducible:

Consistently


Steps to Reproduce:
1. Create a new deployment or scale/refresh an existing one


Actual results:

Deployments take an excessively long time to process and either remain in an incorrect state or fail.


Expected results:

Deployments would work as expected.