[
https://issues.apache.org/jira/browse/OOZIE-3715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17733884#comment-17733884
]
Dénes Bodó commented on OOZIE-3715:
-----------------------------------
[~chenhd]
In TestSignalXCommand please don't use infinite loop:
{code:java}
while (!WorkflowJob.Status.FAILED.equals( engine.getJob(jobId).getStatus() )){
Thread.sleep(500);
} {code}
but invoke *waitFor* method as can be seen in other test cases.
Also please use "uri:oozie:workflow:1.0" instead of "uri:oozie:workflow:0.4" in
the same file and remove space after '(' and before ')'.
Always compare constant to variable:
{code:java}
"somestring".equals(action.getName()) {code}
Isn't there a way to swap
{code:java}
// wait for execute KillXCommand
Thread.sleep(500);{code}
to some safer implementation? Waiting 500 ms may be not enough in some cases so
this could became an intermittently failing test case which we must avoid.
Isn't there an indicator which we could check in a waitFor method like size of
a list of an SQL query executed via JPAExecutor?
Thanks
> Fix fork out more than one transitions submit , one transition submit fail
> can't execute KillXCommand
> -----------------------------------------------------------------------------------------------------
>
> Key: OOZIE-3715
> URL: https://issues.apache.org/jira/browse/OOZIE-3715
> Project: Oozie
> Issue Type: Bug
> Components: core
> Affects Versions: 5.2.1
> Reporter: chenhaodan
> Assignee: chenhaodan
> Priority: Major
> Labels: patch
> Fix For: trunk
>
> Attachments: OOZIE-3715-001.patch, OOZIE-3715-002.patch,
> OOZIE-3715-003.patch, OOZIE-3715-004.patch, OOZIE-3715-005.patch,
> forkSubmitFail_issue.txt, status.png
>
>
> When I fork 2 transitions( A and B) to submit , when A transition failed , B
> transition still Running , because can't execute KillXCommand.
> SignalXCommand.startForkedActions, when one transition submit fail will
> create a new ActionStartXCommand and invoke failJob, failJob will add
> WorkflowNotificationXCommand and KillXCommand to
> {color:#ff0000}*commandQueue*{color} , and callback at XCommand.call method ,
> but we add WorkflowNotificationXCommand and KillXCommand to
> ActionStartXCommand‘s {color:#ff0000}*commandQueue*{color} , but not
> SignalXCommand , so can't execute KillXCommand.
> The code is as follows :
>
> {code:java}
> public void startForkedActions(List<WorkflowActionBean>
> workflowActionBeanListForForked) throws CommandException {
> ......
> for (Future<ActionExecutorContext> result : futures) {
> ......
> if (context.getJobStatus() != null &&
> context.getJobStatus().equals(Job.Status.FAILED)) {
> new ActionStartXCommand(context.getAction().getId(),
> null).failJob(context);
> ......
> }
> ......
> }
> {code}
>
> {code:java}
> public void failJob(ActionExecutor.Context context, WorkflowActionBean
> action) throws CommandException {
> WorkflowJobBean workflow = (WorkflowJobBean) context.getWorkflow();
> if (!handleUserRetry(context, action)) {
> incrActionErrorCounter(action.getType(), "failed", 1);
> LOG.warn("Failing Job due to failed action [{0}]",
> action.getName());
> try {
> workflow.getWorkflowInstance().fail(action.getName());
> WorkflowInstance wfInstance = workflow.getWorkflowInstance();
> ((LiteWorkflowInstance)
> wfInstance).setStatus(WorkflowInstance.Status.FAILED);
> workflow.setWorkflowInstance(wfInstance);
> workflow.setStatus(WorkflowJob.Status.FAILED);
> action.setStatus(WorkflowAction.Status.FAILED);
> action.resetPending();
> queue(new WorkflowNotificationXCommand(workflow, action));
> queue(new KillXCommand(workflow.getId()));
> InstrumentUtils.incrJobCounter(INSTR_FAILED_JOBS_COUNTER_NAME, 1,
> getInstrumentation());
> }
> catch (WorkflowException ex) {
> throw new CommandException(ex);
> }
> }
> }
> {code}
>
> {code:java}
> public final T call() throws CommandException {
> if (commandQueue != null) {
> for (Map.Entry<Long, List<XCommand<?>>> entry :
> commandQueue.entrySet()) {
> LOG.debug("Queuing [{0}] commands with delay [{1}]ms",
> entry.getValue().size(), entry.getKey());
> if (!callableQueueService.queueSerial(entry.getValue(),
> entry.getKey())) {
> LOG.warn("Could not queue [{0}] commands with delay [{1}]ms,
> queue full", entry.getValue()
> .size(), entry.getKey());
> }
> }
> }
> }
> {code}
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
