Details
-
Task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
1.5.0
-
None
Description
Steps to reproduce:
- Start 10 node cluster.
- On Step-9 of installer wizard while ongoing install command, Stopped ambari-agent on 9 hosts.
- Hit Retry button.
- Some of the host_components ended up in INSTALL_FAILED state, some in UNKNOWN state and others in INSTALLING state. Host components in INSTALLING state never transitioned to any other state.
Reason:
The ActionScehduler.doWork() depends on the HostRoles to timeout before aborting the operation in order to update Host Component status.
This particular issue is caused by a the timing of the change in Host state to UNKNOWN, which happens after processing of in progress stages and before Abort.
Attachments
Issue Links
- links to