Dmytro Vitiuk created AMBARI-25606:
--------------------------------------
Summary: Sometimes request aborting doesn't abort IN_PROGRESS task
Key: AMBARI-25606
URL: https://issues.apache.org/jira/browse/AMBARI-25606
Project: Ambari
Issue Type: Bug
Components: ambari-server
Affects Versions: 2.7.5
Reporter: Dmytro Vitiuk
Fix For: 2.7.6
There is a raise condition in request aborting process.
Imagine, we have a request with couple of stages/tasks and one task is in
IN_PROGRESS state. Then let's try to abort the request.
Crucial steps of request aborting are:
# Get all not completed tasks for the request.
# Get QUEUED and IN_PROGRESS tasks from #1 and send cancel commands to the
agents (we've sent cancel command for our task).
# Get tasks in HOLDING_STATES (HOLDING_FAILED etc.) from #1 and try to abort
them.
# Get all tasks form stages which are not in COMPLETED state for the request.
# Get tasks from #4 which are in PENDING, QUEUED, IN_PROGRESS states and abort
them.
But if the server receive response from the agent for cancel command (sent in
#2), then task will be transitioned from IN_PROGRESS to HOLDING_FAILED state.
That's ok if it happen before step #2 (it will be aborted because has
HOLDING_FAILED state) or after step #5 (it had IN_PROGRESS state and was
aborted; update from the agent won't change state to HOLDING_FAILED, because we
handle this case).
But if we receive response from the agent between steps #3 and #5, then task
will not be aborted and we will get a task with HOLDING_FAILED state after
abort operation completion.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
