'Hiding' Rosetta/BOINC?

Message boards : Number crunching : 'Hiding' Rosetta/BOINC?

To post messages, you must log in.

AuthorMessage
Zifnab

Send message
Joined: 25 Mar 07
Posts: 8
Credit: 6,369
RAC: 0
Message 38391 - Posted: 25 Mar 2007, 22:44:36 UTC

I've recently received permission to run Rosetta across two labs of Macs that I administer, but we don't want the BOINC application to have to show up in the dock. Having it in the bar up by the clock is fine, but we want to prevent users from fiddling with it, essentially. Is there a way to run it without the application showing in the dock?
ID: 38391 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 38402 - Posted: 26 Mar 2007, 0:39:15 UTC

Maybe you should concentrate on getting one machine working before you try it on two labs of machines.

It appears you haven't been able to return a single valid result yet...
Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 38402 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Zifnab

Send message
Joined: 25 Mar 07
Posts: 8
Credit: 6,369
RAC: 0
Message 38453 - Posted: 27 Mar 2007, 0:34:16 UTC - in response to Message 38402.  

Mainly due to our desire to have the process as invisible to the user as possible. It runs, it downloads, and it processes, and I don't see what my number of returned results has to do with the validity of my question...
ID: 38453 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 38456 - Posted: 27 Mar 2007, 1:27:33 UTC

Two things:

You asking a BOINC core client/manager question on a project forum. The BOINC related questions are better asked in the BOINC forums.

My concern about the failed WUs is that I could imagine 2 labs full of MACs all downloading and burning through WUs and producing nothing, if they all had the same problem as your own box.


Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 38456 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Zifnab

Send message
Joined: 25 Mar 07
Posts: 8
Credit: 6,369
RAC: 0
Message 38457 - Posted: 27 Mar 2007, 1:48:00 UTC
Last modified: 27 Mar 2007, 1:51:57 UTC

Ah, my apologies for asking this in the wrong place. I didn't realize there was a better place for my question.


Edit: Wow, I just looked at my results for the first time. I didn't realize I was erroring on so many WUs. Any idea what's going on?
ID: 38457 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 38458 - Posted: 27 Mar 2007, 2:11:20 UTC
Last modified: 27 Mar 2007, 2:13:05 UTC

Nope - I have no experience with the MAC client or OS, but you're asking that question in the right place!

There is something in your WU error messages about keeping in memory
<stderr_txt>
Too many restarts with no progress. Keep application in memory while preempted.


Are you letting the WU run un-interrupted? Running other BOINC projects with the task switching set too low?

Evidently the current client doesn't checkpoint within a model, so if the first model never finishes, you just keep starting over again until the client says "enough!" and sends it back.
Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 38458 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile paulcsteiner

Send message
Joined: 15 Oct 05
Posts: 19
Credit: 3,144,322
RAC: 542
Message 38460 - Posted: 27 Mar 2007, 2:30:55 UTC - in response to Message 38457.  

Ah, my apologies for asking this in the wrong place. I didn't realize there was a better place for my question.


Edit: Wow, I just looked at my results for the first time. I didn't realize I was erroring on so many WUs. Any idea what's going on?


Hi Zifnab, Not necessarily the wrong place to post.
These types of errors are sometimes caused by repeated interruption of the client.
Although because you've got so many in a short time span with machine #448935 I'm thinking maybe a possible thermal issue? (the Rosetta cognoscenti will post here soon to clarify)Or is there some other process running thats pegged out the machine?
Also, for the same machine, if you look at the error message you will see the following text: :Too many restarts with no progress. Keep application in memory while preempted."
I believe the second bit can be solved by changing your "General preferences" for BOINC to YES for this option "Leave applications in memory while suspended?
(suspended applications will consume swap space if 'yes')"



ID: 38460 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Zifnab

Send message
Joined: 25 Mar 07
Posts: 8
Credit: 6,369
RAC: 0
Message 38462 - Posted: 27 Mar 2007, 3:15:23 UTC - in response to Message 38460.  

I don't think there's a thermal issue at fault here. These machines have sustained an extremely high processing load for very long periods before without issue, so it seems odd that a thermal or other issue would rear its head in the first 30 seconds that I have the client running.

I changed the memory setting that you suggested to YES, but that doesn't appear to have fixed the issue. I actually went ahead and started a new thread about this, with a more descriptive title and OP. I think the issue is related to the constant "Process XXXX not found" errors that I'm receiving, but I don't know what the cause of them would be.
ID: 38462 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 38463 - Posted: 27 Mar 2007, 3:20:26 UTC

Zifnab, your question is welcome here, as well as on the BOINC boards mentioned earlier. I am not familiar with Mac either, so couldn't offer an answer to your specific question.

As for tasks being ended due to too many restarts, this is one of Rosetta's failsafe mechanisms that helps assure things keep moving without intervention. Each time a task is restarted from a checkpoint, a check is made to see if we've begun from this same place before and a count is kept of how many times this occurs. Once you restart from the same checkpoint without making progress 5 times, then the WU is marked as completed.

What might cause a task to be restarted so many times?

BOINC is trying to manage a number of resources and meet a number of goals. Your system has 2 CPUs, so depending upon how much memory you allow BOINC to use, it may begin a task, and then as it gets up to speed and consumes more memory, BOINC may decide it is taking too much and suspend the task. It then begins another task, hoping to find one that will run with less memory. Once another task completes, it's BOINC memory is freed up and so it may again take on one that it had suspended previously.

As mentioned in the message, and previously in this thread, if your General Preference is set to NOT keep in memory when suspended, then every time BOINC begins the task, and it begins to use more memory, it then suspends it and goes back to the beginning, because you didn't get far enough along to reach a checkpoint.

Another factor with that is whether you allow BOINC to run while the machine is in use. Unless you are really using intensive graphics or using lots of memory, most people find BOINC can run all the time, in use or not, and not have it impact their applications. This is because BOINC runs at a low priority. Anyway, regardless of your preference, just be aware that if you do not keep applications in memory, then every time the machine goes idle, and start crunching BOINC and then you come back to use it, you're basically throwing away whatever work was done during the half hour or whatever that your machine was idle. If you machine do not have a long stretch of time being idle, and you do not keep applications in memory when preempted, and you do not allow BOINC to continue running while the computer is in use, that might explain what you are seeing. And we don't want to change when you use the computer, so change one or both of the other settings and see if your machine is still responsive with the applications you wish to use.

If you exit BOINC, the tasks have to restart from their last checkpoint.

If you turn off your PC, the tasks have to restart from their last checkpoint.

And these are normal things, and that is why if they happen just 2 or 3 times we keep chugging. But 5 times in a row with no progress, generally means this task is not well suited to the environment it is trying to run in, and it would be better to report back what you've got and get another task which may run better. That's the idea anyway. But it's certainly not supposed to happen every time.

So, three things to check, your % memory you allow BOINC to use, both while active and while inactive, and you want to set the leave applications in memory when preempted to yes... unless you have some specific need not to do so (memory intensive application you use a lot for example). And thirdly, whether you allow BOINC to run while the machine is in use. If you pop in and out of the machine periodically, and do NOT keep the applications in memory while preempted, then you are only really going to get work done at night if you leave your machine on and idle for hours at a time.

I've added a post on your behalf in the Problems with... thread.

If you could add to the Problems with... thread some comments about whether you are running more projects then just Rosetta, and when you review the tasks in the BOINC Manager, what is shown for the status of the tasks, that might be helpful. Do the tasks show "waiting for memory"?

Your tasks show 30 nstructs compelted, but zero decoys. So it seems some useful for was done. But, that's not how the results would normally appear.
Rosetta Moderator: Mod.Sense
ID: 38463 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Conan
Avatar

Send message
Joined: 11 Oct 05
Posts: 150
Credit: 3,858,040
RAC: 1,584
Message 38484 - Posted: 27 Mar 2007, 12:14:14 UTC - in response to Message 38391.  

I've recently received permission to run Rosetta across two labs of Macs that I administer, but we don't want the BOINC application to have to show up in the dock. Having it in the bar up by the clock is fine, but we want to prevent users from fiddling with it, essentially. Is there a way to run it without the application showing in the dock?


The only way I know is to install as a Service. It then does not show at all on the computer desktop. With Windows you need administrator rights, unsure on Linux (which I use but the computers don't do anything but Boinc so no need for Service install), same with Mac's, I am unsure but should work.

It seems like a dream to get one more computer let alone two labs worth.
ID: 38484 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Samuel

Send message
Joined: 28 Mar 07
Posts: 1
Credit: 19,011
RAC: 0
Message 38531 - Posted: 28 Mar 2007, 17:39:35 UTC

You could run the command line client, and as a daemon.

See this: http://boinc.berkeley.edu/mac_admin_tools.php
ID: 38531 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : 'Hiding' Rosetta/BOINC?



©2024 University of Washington
https://www.bakerlab.org