Meng-Shuan Tsai created HDDS-16194:
--------------------------------------
Summary: OM decommission fails after scale-down below Ratis
majority
Key: HDDS-16194
URL: https://issues.apache.org/jira/browse/HDDS-16194
Project: Apache Ozone
Issue Type: Sub-task
Components: Helm, kubernetes
Reporter: Meng-Shuan Tsai
Assignee: Meng-Shuan Tsai
h3. Problem
With {{{}om.persistence.enabled=true{}}}, OM scale-down can fail in the
post-upgrade decommission Job because Helm applies the smaller StatefulSet
*before* that hook.
h3. Steps to reproduce (kind)
# Persistence on for om/scm/datanode. Control: {{5 -> 3}} — both decommission
Jobs succeed, roles=3.
# Failure: {{3 -> 1}} — STS {{{}1/1{}}}; decommission of {{…-om-1}} fails.
h3. Actual behavior
{code:java}
OM:ozone-om-0 is not the leader. Could not determine the leader node
{code}
No success / wipe; removed ordinals' PVCs stay Bound with old data.
h3. Expected behavior
Decommission succeeds, or the chart refuses unsafe scale-down. Ring matches
{{{}om.replicas{}}}.
h3. Analysis
{{pre-upgrade}} → apply STS → {{post-upgrade}} decommission. RemoveOM needs
majority of the *previous* config: {{5 -> 3}} live 3 ≥ 3 works; {{3 -> 1}} live
1 < 2 fails. Failed Job skips {{{}rm -rf{}}}.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
