It looks weird to me too. At every start progress is at 2% and goes to 11% in 60 minutes.
Tullio
In fact, it's arrive to me too.
The task will finally ok but I lost 5 hours of CPU time.
It's for that I write the number'n'crunching message for E@H
It's seems that when "Global correlations S5" is stopped it's lost data and recomputing
It's Happen particulary on cosmology@home project
be careful of your RAM capacities
estimated time is often false on BOINC. For me it's return normal when the task runs correctly
Each time the computer is restart, the progress falls back tens of percent. This happens only to S5GC1, ABP1 is OK. The WU 74128878 needed 34,5 hours. Normal, I need 13 hours.
I have a second Linux machine in which this happens. The WUs are still ongoing. Do you need any log files?
A workaround might be to allow suspended tasks to stay in memory.
If that's a workaround, the cause might be the new preference option to suspend tasks if non-BOINC activity exceeds a certain percentage (the default being 25%). That could have the same effect as the BOINC CPU-throttling, namely stopping and starting the running tasks every few seconds.
Alas, there's no entry for that option at the Einstein@Home (or CPDN) computing preferences yet:
Suspend work if CPU usage is above 25 %
0 means no restriction
Enforced by version 6.10.30+
So, users who aren't attached to another project with newer server software (of "mine" that's SETI@home and The Lattice Project) need to use the local preferences to set that option to zero, with the drawback that all future modifications must be done locally also, as the local preferences (as a whole) override the online ones.
I'm not sure if account managers (BAM!, GridRepublic, ...) support that preference already.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
I am using only English to
)
I am using only English to speak to a computer. Should I diminish the 60 seconds value?
Tullio
RE: I am using only English
)
60 seconds is a good value, but the WU uses the 300 seconds setting from "somewhere else " (other project or local settings in Boinc manager).
Even 300 seconds should be more or less ok, so I wonder why the app makes so little progress over time.
Can you do a
grep checkpoint stderr.txt in tghe slot subdir of the task? Just to check it is able to read the checkpoints at all when restarting.
CU
HBE
Here it is: 2010-05-12
)
Here it is:
2010-05-12 17:35:31.9175 (3080) [normal]: INFO: No checkpoint h1_0550.35_S5R4__141_S5GC1a_1_0.cpt found - starting from scratch
2010-05-12 20:02:41.8759 (3722) [debug]: Successfully read checkpoint:9
2010-05-12 22:10:23.5354 (4442) [debug]: Successfully read checkpoint:8
2010-05-13 00:56:41.7095 (4688) [debug]: Successfully read checkpoint:8
2010-05-13 09:59:28.4771 (9057) [debug]: Successfully read checkpoint:8
2010-05-13 12:49:03.3680 (11332) [debug]: Successfully read checkpoint:24
2010-05-13 14:55:14.0092 (12386) [debug]: Successfully read checkpoint:16
2010-05-13 18:28:57.7389 (14493) [debug]: Successfully read checkpoint:15
2010-05-13 21:51:17.1074 (15939) [debug]: Successfully read checkpoint:8
2010-05-13 23:51:53.7125 (16110) [debug]: Successfully read checkpoint:8
2010-05-14 02:37:58.7707 (16350) [debug]: Successfully read checkpoint:9
2010-05-14 04:43:41.6052 (16523) [debug]: Successfully read checkpoint:8
2010-05-14 12:04:53.5782 (18169) [debug]: Successfully read checkpoint:11
2010-05-14 14:34:56.0849 (18505) [debug]: Successfully read checkpoint:15
2010-05-14 17:50:48.8759 (19819) [debug]: Successfully read checkpoint:15
2010-05-14 21:41:20.6521 (21141) [debug]: Successfully read checkpoint:8
2010-05-14 23:54:56.1304 (21364) [debug]: Successfully read checkpoint:9
2010-05-15 02:43:16.3771 (21600) [debug]: Successfully read checkpoint:9
2010-05-15 08:27:25.2248 (22335) [debug]: Successfully read checkpoint:9
2010-05-15 10:00:48.4263 (22966) [debug]: Successfully read checkpoint:15
2010-05-15 13:29:58.1347 (24164) [debug]: Successfully read checkpoint:15
This looks weird...I'll ask
)
This looks weird...I'll ask the devs if this is supposed to be this way...
Thanks for providing the logs
CU
HBE
It looks weird to me too. At
)
It looks weird to me too. At every start progress is at 2% and goes to 11% in 60 minutes.
Tullio
RE: It looks weird to me
)
In fact, it's arrive to me too.
The task will finally ok but I lost 5 hours of CPU time.
It's for that I write the number'n'crunching message for E@H
It's seems that when "Global correlations S5" is stopped it's lost data and recomputing
It's Happen particulary on cosmology@home project
be careful of your RAM capacities
estimated time is often false on BOINC. For me it's return normal when the task runs correctly
just a poet
The devs will be looking into
)
The devs will be looking into it.
A workaround might be to allow suspended tasks to stay in memory.
CU
HBE
Each time the computer is
)
Each time the computer is restart, the progress falls back tens of percent. This happens only to S5GC1, ABP1 is OK. The WU 74128878 needed 34,5 hours. Normal, I need 13 hours.
I have a second Linux machine in which this happens. The WUs are still ongoing. Do you need any log files?
Bye, Grubix.
RE: The devs will be
)
I'll try this workaround and report if things go better.
Tullio
RE: A workaround might be
)
If that's a workaround, the cause might be the new preference option to suspend tasks if non-BOINC activity exceeds a certain percentage (the default being 25%). That could have the same effect as the BOINC CPU-throttling, namely stopping and starting the running tasks every few seconds.
Alas, there's no entry for that option at the Einstein@Home (or CPDN) computing preferences yet:
So, users who aren't attached to another project with newer server software (of "mine" that's SETI@home and The Lattice Project) need to use the local preferences to set that option to zero, with the drawback that all future modifications must be done locally also, as the local preferences (as a whole) override the online ones.
I'm not sure if account managers (BAM!, GridRepublic, ...) support that preference already.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
