Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 367745

Summary:

Zombie operations in Orion started after a week

Product:

[ECD] Orion

Reporter:

Tomasz Zarna <tomasz.zarna>

Component:

Server

Assignee:

Malgorzata Janczarska <malgorzata.tomczyk>

Status:

RESOLVED FIXED

QA Contact:

Severity:

minor

Priority:

Version:

0.4

Target Milestone:

0.4

Hardware:

OS:

Windows XP

Whiteboard:

Attachments:

Description	Flags
Screenshot showing zombie ops	none
Screenshot showing my All Operations page	none

Description Tomasz Zarna

2012-01-03 06:38:21 EST

I've started Orion server this morning after a long xmas break. To my surprise, when I opened Git Status for a project I saw two ops running in the Recent operations popup:
* mysterious "Pulling..."
* and "Pulling gotom..."

(see the screenshot)

They were probably started days ago and failed. They should not be shown here and now.

Comment 1 Tomasz Zarna

2012-01-03 06:39:50 EST

Created attachment 208930 [details]
Screenshot showing zombie ops

Comment 2 Malgorzata Janczarska

2012-01-03 06:45:56 EST

Is this by any chance a dup of Bug 367726? Could it happen that you restarted your server while pulling was running?
Go to "All Operations" and check the schedule date to know the exact date.

Comment 3 Tomasz Zarna

2012-01-03 06:55:47 EST

Sorry, I don't remember if I restarted the server back then. So the exact date would tell me nothing.

However, there some interesting details on the Operations page:
* The "Pulling gotom" op seems to failed because of connection problem : "git://github.com/gotom/gotom.git: unknown host". No scheduled date.
* The other one, "Pulling" has "Auth failed" status. Again, no scheduled date.

Comment 4 Malgorzata Janczarska

2012-01-03 07:04:17 EST

(In reply to comment #3)
> However, there some interesting details on the Operations page:
> * The "Pulling gotom" op seems to failed because of connection problem :
> "git://github.com/gotom/gotom.git: unknown host". No scheduled date.
> * The other one, "Pulling" has "Auth failed" status. Again, no scheduled date.

So you are telling that on operations page they are not marked as running?

Comment 5 Tomasz Zarna

2012-01-03 07:11:49 EST

Created attachment 208932 [details]
Screenshot showing my All Operations page

Comment 6 Malgorzata Janczarska

2012-01-03 07:40:47 EST

Ok, I can see it now. It seems that pull job has done something nasty: the task has failed, but instead of marking existing task as failed it created a new one. This is only a guess, but operations in the list are sorted by last update.
This is OK if they don't have schedule date, I added this attribute some time ago, and probably they where schedule before it.
I'll have to look at the pull job to find out what might cause it.

Comment 7 Tomasz Zarna

2012-01-03 10:29:57 EST

To make things worse, Client keeps asking for these tasks as they are both "Running" with "PercentComplete" equal 0 :|

Comment 8 Malgorzata Janczarska

2012-01-16 04:30:35 EST

I'm working on it along with Bug 364831. When I migrate all tasks to TaskJob we will have greater control over task lifecycle and if job is finished and task is not marked as completed it'll just change to fail.

>To make things worse, Client keeps asking for these tasks as they are both
>"Running" with "PercentComplete" equal 0 :|
Client will always ask for the task status when it's marked as running, so all we can fix if server. Unless you have an idea how could client know that he should stop tracking the task.

Comment 9 Malgorzata Janczarska

2012-01-23 05:16:18 EST

In Bug 364831 I migrated git tasks to TaskJob. This should solve the problem of never ending tasks.