Orphaned BOINC process

Message boards : Number crunching : Orphaned BOINC process
Message board moderation

To post messages, you must log in.

AuthorMessage
j2satx
Volunteer tester

Send message
Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
United States
Message 135131 - Posted: 10 Jul 2005, 14:46:46 UTC


Using BOINC Manager in a "multiple" computer environment can cause an "orphaned" BOINC process.

I have not found a way to regain control of the "orphaned" process, other than to re-boot the computer the process is on.

I can "stop" the task using one of my utilities, but it will cause an error in the WU.

It seems that some of the time I can "end task" with the Windows Task Manager and the WU will not be damaged.

There is another thread where two BOINC processes seem to be running simultaneously on an AMD. My orphaned processes do not run simultaneously, so this may not have any relation to that issue.

Has anyone else experienced this "orphan" issue?


ID: 135131 · Report as offensive
Profile Paul D. Buck
Volunteer tester

Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 135143 - Posted: 10 Jul 2005, 15:21:33 UTC

Yes. Known problem, my experience is that it is more common under Unix varients (OS-X in my case) than Windows. Cause is unknown ...

Killing the process has a 50/50 chance of destroying the result, again in my experience. rebooting is usually safest ...
ID: 135143 · Report as offensive
j2satx
Volunteer tester

Send message
Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
United States
Message 135209 - Posted: 10 Jul 2005, 17:58:02 UTC
Last modified: 10 Jul 2005, 18:02:22 UTC

In Windows I can cause it to happen anytime. If you open the manager (local host) and select another computer and then close the manager......the local host process, started when you opened manager, becomes orphaned.

I would think a cure could be to not have manager automatically open a process on local host, but allow an option to start one.
ID: 135209 · Report as offensive
Profile ralic
Volunteer tester

Send message
Joined: 6 Jan 00
Posts: 308
Credit: 274,230
RAC: 0
Message 135503 - Posted: 11 Jul 2005, 9:25:52 UTC - in response to Message 135209.  

In Windows I can cause it to happen anytime. If you open the manager (local host) and select another computer and then close the manager......the local host process, started when you opened manager, becomes orphaned.

I would think a cure could be to not have manager automatically open a process on local host, but allow an option to start one.

I saw this one again myself recently. The problem was fixed way back in v4.22 but appears to have resurfaced. I have already advised the alpha crew and re-opend the bug on BoincZilla Bug #3

Thanks anyway for your report.
ID: 135503 · Report as offensive
j2satx
Volunteer tester

Send message
Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
United States
Message 135584 - Posted: 11 Jul 2005, 15:01:31 UTC - in response to Message 135503.  

In Windows I can cause it to happen anytime. If you open the manager (local host) and select another computer and then close the manager......the local host process, started when you opened manager, becomes orphaned.

I would think a cure could be to not have manager automatically open a process on local host, but allow an option to start one.

I saw this one again myself recently. The problem was fixed way back in v4.22 but appears to have resurfaced. I have already advised the alpha crew and re-opend the bug on BoincZilla Bug #3

Thanks anyway for your report.


I appreciate the response, thanks.
ID: 135584 · Report as offensive
Profile ralic
Volunteer tester

Send message
Joined: 6 Jan 00
Posts: 308
Credit: 274,230
RAC: 0
Message 135602 - Posted: 11 Jul 2005, 16:07:03 UTC - in response to Message 135584.  
Last modified: 11 Jul 2005, 16:10:02 UTC

I appreciate the response, thanks.

You're welcome. :)
I have not found a way to regain control of the "orphaned" process, other than to re-boot the computer the process is on.

I can "stop" the task using one of my utilities, but it will cause an error in the WU.

It seems that some of the time I can "end task" with the Windows Task Manager and the WU will not be damaged.

To regain some degree of control of the orphan, just start BOINC Manager on the same host. It should find the running boinc.exe process and attach to it. If not, a "File->Select Computer...->OK" should re-attach.

Once attached, suspend BOINC (File->Suspend), wait for several seconds to give BOINC a chance to suspend any running science apps (watch your processor usage in task manager), then kill the process as you like. This will give you a better chance of saving your wu's from any errors.

After you've confirmed the process to be dead, restart BOINC and resume (File->Run based on preferences or File->Run always).

[edited for spelling and readability]
ID: 135602 · Report as offensive
j2satx
Volunteer tester

Send message
Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
United States
Message 135662 - Posted: 11 Jul 2005, 19:23:13 UTC


I followed your process and it is true that you can suspend the running process, but killing it makes it error.

Can the manager be programmed so it does not automatically start a process when it is opened?

Can the manager be programmed so you will receive a warning that you are very likely to create an orphan if you have opened a process and are going to shut down manager when you have selected any host other than "local host"?

Thank you.
ID: 135662 · Report as offensive
Profile ralic
Volunteer tester

Send message
Joined: 6 Jan 00
Posts: 308
Credit: 274,230
RAC: 0
Message 135955 - Posted: 13 Jul 2005, 7:46:51 UTC - in response to Message 135662.  

I followed your process and it is true that you can suspend the running process, but killing it makes it error.

Not quite sure I'm following 100%. Just to clarify, when you kill the process after suspending boinc, does the wu still error? If so, for information sake, which project is behaving like this?
Can the manager be programmed so it does not automatically start a process when it is opened?

It could be, but doing so would make boinc a little more complicated for the average user. When installed in single-user or shared mode, it is the manager that starts boinc when launched and stops boinc when exited. Changing this behaviour would only result in having to find some other way of starting and stopping the boinc process for single and shared users.
Can the manager be programmed so you will receive a warning that you are very likely to create an orphan if you have opened a process and are going to shut down manager when you have selected any host other than "local host"?

Well, the manager is actually not supposed to create these "orphans". This is where the bug comes in.
The proper (and previous) functioning of the manager is basically as follows:
1.) When launched, start a boinc process if one is not already running.
2.) If a boinc process was started by the manager, shut it down when exiting.
3.) When connecting to a remote computer, never shut down the remote boinc process when exiting.

The bug lies in items 2 and 3 above.
Currently item 2 fails if the manager is connected to a remote host when exited and item 3 occurs when it should never do so.

This incorrect behaviour was previously seen and corrected. Unfortunately it has resurfaced and the devs need to isolate it and fix it.
ID: 135955 · Report as offensive
j2satx
Volunteer tester

Send message
Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
United States
Message 136041 - Posted: 13 Jul 2005, 13:52:01 UTC - in response to Message 135955.  

[quote]I followed your process and it is true that you can suspend the running process, but killing it makes it error.

Not quite sure I'm following 100%. Just to clarify, when you kill the process after suspending boinc, does the wu still error? If so, for information sake, which project is behaving like this?

Predictor was the last one that errored after suspending. SETI and SETI Beta errored previously, but I did not try them again by suspending the project first.
ID: 136041 · Report as offensive
j2satx
Volunteer tester

Send message
Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
United States
Message 136042 - Posted: 13 Jul 2005, 13:55:22 UTC - in response to Message 135955.  

Can the manager be programmed so it does not automatically start a process when it is opened?

It could be, but doing so would make boinc a little more complicated for the average user. When installed in single-user or shared mode, it is the manager that starts boinc when launched and stops boinc when exited. Changing this behaviour would only result in having to find some other way of starting and stopping the boinc process for single and shared users.
Can the manager be programmed so you will receive a warning that you are very likely to create an orphan if you have opened a process and are going to shut down manager when you have selected any host other than "local host"?


I had to change strategy and start running boinc.exe first. When I want to look at something, I open boincmgr and then close it when finished.

I am not willing to use boincmgr continuously just to run boinc processes when it consumes 3% of my faster CPUs and 10%+ of my slower one.
ID: 136042 · Report as offensive

Message boards : Number crunching : Orphaned BOINC process


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.