Welcome to MilkyWay@home

Lost WU

Message boards : Number crunching : Lost WU
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Fran

Send message
Joined: 3 Sep 07
Posts: 2
Credit: 439,099
RAC: 0
Message 4149 - Posted: 15 Jul 2008, 19:43:20 UTC
Last modified: 15 Jul 2008, 19:46:08 UTC

I finished computation of wu gs_3730382_1216077095_2569343_0 but it is not in my list. Seems like it never been in list...
How it is possible?

ID Date Project Message
48 15.7.2008 21:20:47 Milkyway@home Computation for task gs_3730382_1216077095_2569343_0 finished
49 15.7.2008 21:20:47 Milkyway@home Starting gs_3721282_1216047814_2550830_1
50 15.7.2008 21:20:47 Milkyway@home Starting task gs_3721282_1216047814_2550830_1 using astronomy version 122
51 15.7.2008 21:20:49 Milkyway@home [file_xfer] Started upload of file gs_3730382_1216077095_2569343_0_0
52 15.7.2008 21:20:55 Milkyway@home [file_xfer] Finished upload of file gs_3730382_1216077095_2569343_0_0
53 15.7.2008 21:20:55 Milkyway@home [file_xfer] Throughput 476 bytes/sec
54 15.7.2008 21:21:59 --- Time passed...reporting result now.
55 15.7.2008 21:21:59 Milkyway@home Sending scheduler request: To report completed tasks
56 15.7.2008 21:21:59 Milkyway@home Reporting 1 tasks
57 15.7.2008 21:22:04 Milkyway@home Scheduler RPC succeeded [server version 511]
ID: 4149 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Alinator

Send message
Joined: 7 Jun 08
Posts: 464
Credit: 56,639,936
RAC: 0
Message 4150 - Posted: 15 Jul 2008, 19:49:28 UTC - in response to Message 4149.  
Last modified: 15 Jul 2008, 19:49:55 UTC

I finished computation of wu gs_3730382_1216077095_2569343_0 but it is not in my list. Seems like it never been in list...
How it is possible?
I have 4 wus in the list and they are now in my computer but the one above is not.

ID Date Project Message
48 15.7.2008 21:20:47 Milkyway@home Computation for task gs_3730382_1216077095_2569343_0 finished
49 15.7.2008 21:20:47 Milkyway@home Starting gs_3721282_1216047814_2550830_1
50 15.7.2008 21:20:47 Milkyway@home Starting task gs_3721282_1216047814_2550830_1 using astronomy version 122
51 15.7.2008 21:20:49 Milkyway@home [file_xfer] Started upload of file gs_3730382_1216077095_2569343_0_0
52 15.7.2008 21:20:55 Milkyway@home [file_xfer] Finished upload of file gs_3730382_1216077095_2569343_0_0
53 15.7.2008 21:20:55 Milkyway@home [file_xfer] Throughput 476 bytes/sec
54 15.7.2008 21:21:59 --- Time passed...reporting result now.
55 15.7.2008 21:21:59 Milkyway@home Sending scheduler request: To report completed tasks
56 15.7.2008 21:21:59 Milkyway@home Reporting 1 tasks
57 15.7.2008 21:22:04 Milkyway@home Scheduler RPC succeeded [server version 511]


This is most likely due to the Task Purging Daemon being set to get rid of completed tasks almost as soon as they pass validation and get assimilated.

IOW's, if you aren't there to take a look at the Host Summary page within a few minutes at most of when the task gets reported, you most likely aren't going to see it.

Hopefully, now that the heat seems to be off the backend with the new longer running work, they can set the daemon to let the completed tasks hang around a bit more. ;-)

HTH,

Alinator
ID: 4150 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Fran

Send message
Joined: 3 Sep 07
Posts: 2
Credit: 439,099
RAC: 0
Message 4151 - Posted: 15 Jul 2008, 20:02:14 UTC

Thanks a lot. It does make sense. But I think, this is not good, because I don`t know, if the wu was validated and if I got credit and how much.
ID: 4151 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Alinator

Send message
Joined: 7 Jun 08
Posts: 464
Credit: 56,639,936
RAC: 0
Message 4152 - Posted: 15 Jul 2008, 21:28:25 UTC - in response to Message 4151.  
Last modified: 15 Jul 2008, 21:29:31 UTC

Thanks a lot. It does make sense. But I think, this is not good, because I don`t know, if the wu was validated and if I got credit and how much.


Agreed, but the project team really kind of had their back to the wall over the last couple of weeks with the Database Server grinding to a halt due to the volume of service requests made on it. It had gotten to the point where the project was coming to virtual standstill every couple of days or so, and something had to be done. ;-)

Until they get the daemon set to delay purging for more than ASAP, the only workaround I know of is to set BOINC to disable network access, and then report 'manually' when you are there to monitor the Host Summary. Be advised that this isn't quite as simple to do as it sounds, and is pretty much the same PITA as having the completed tasks go poof instantly if you like tracking your hosts yourself. :-)

Alinator
ID: 4152 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Conan
Avatar

Send message
Joined: 2 Jan 08
Posts: 122
Credit: 69,479,692
RAC: 1,456
Message 4209 - Posted: 17 Jul 2008, 13:08:38 UTC - in response to Message 4151.  

Thanks a lot. It does make sense. But I think, this is not good, because I don`t know, if the wu was validated and if I got credit and how much.


@Fran, it seems you get about 5 minutes, then results are purged from the database.
So upload, then quickly check the results or you miss out.
ID: 4209 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Alinator

Send message
Joined: 7 Jun 08
Posts: 464
Credit: 56,639,936
RAC: 0
Message 4211 - Posted: 17 Jul 2008, 16:35:25 UTC - in response to Message 4209.  
Last modified: 17 Jul 2008, 16:36:29 UTC

Thanks a lot. It does make sense. But I think, this is not good, because I don`t know, if the wu was validated and if I got credit and how much.


@Fran, it seems you get about 5 minutes, then results are purged from the database.
So upload, then quickly check the results or you miss out.


If that, I was just logging completed task data for my hosts and the completed ones got purged within 2 to 3 minutes. I agree though that 5 minutes is probably pretty close to the maximum time they hang around right at the moment.

So I would guess the time depends on where they end up in the purge queue, and where the daemon happens to be in the list as it works through it.

Just for reference, I was talking with Travis about this a few days ago and the problem for the project was that the BOINC Server software package only allows the admins to set the purge delay in terms of days (0 to whatever you want). However, with the old work this was causing a problem because the database table of completed results would get so big that the purge daemon would start burning up so much time all the other processes ground to a virtual standstill fairly quickly.

I would think since the runtimes have increase at least a factor of 30 or so, they might be able to get away with the one day setting. In any event, Travis said he was going to look into modifying the code so he could have resolution in hours rather than days when he gets a chance, but I would imagine he has his hands full right at the moment getting the run parameters zeroed in for the new work.

HTH,

Alinator
ID: 4211 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Keck_Komputers
Avatar

Send message
Joined: 29 Aug 07
Posts: 20
Credit: 2,710,089
RAC: 0
Message 4231 - Posted: 18 Jul 2008, 7:31:07 UTC

I thought that "day" was a float, so you can put 0.1 for a little over 2 hours.
BOINC WIKI

BOINCing since 2002/12/8
ID: 4231 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Alinator

Send message
Joined: 7 Jun 08
Posts: 464
Credit: 56,639,936
RAC: 0
Message 4238 - Posted: 18 Jul 2008, 16:17:32 UTC

Apparently not, since when I suggested that Travis said he already tried making the purge delay less than 1 day but it didn't seem to work. I did check the official BOINC documentation, and it doesn't say one way or the other.

I agree though that it doesn't seem to make much sense to have this parameter cast as an integer, but I didn't go rooting through the source to verify it and it wouldn't be the first time that something 'weird' was done in the BOINC code.

I suppose one could make the argument that the only reason for the delay is so users can take a look at what their hosts are doing if they want to. So a delay of less than 24 hours wouldn't be much help there and multiples of 1 day was sufficient if a project wanted to let them hang around longer, since most folks don't sit at the console 24/7 waiting for tasks to show up on Host Summary. ;-)

Alinator
ID: 4238 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Lost WU

©2024 Astroinformatics Group