Collateral damage caused by SETI server hang :(
21 Feb 2005 10:49:28 UTC
Topic 187804
(moderation:
As the title states there is a very bad bug in the BOINC CC:
Since the SETI server nearly constantly ignores requests for work at the moment my clients start to suck up units from Einstein (at least one Einstein unit for every unresponded request to SETI). The queue with Einstein units is now nearly too big for the completion time. When SETI comes back to life the client will suck SETI units and then won't be able to complete the Einstein units in time.
This is a big bug in the concept! :(
Can Einstein somehow slow down giving out work to clients, that aren't able to finish the work in time?
Aloha, Uli
Collateral damage caused by SETI server hang :(
)
No problem with ver 4.22 in this regard.......
Link to Unofficial Wiki for BOINC, by Paul and Friends
If i were you i would set the
)
If i were you i would set the “connect to project every� value to something lower (maybe half) till SETI is up and running again since you have this problem. It would be interesting to se if you would get more WU’s after that.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
Your computers are hidden so
)
Your computers are hidden so I can't see what CC version you are using. I'm running 4.19 and haven't been getting excess WUs on any project due to S@H's problems.
Has the request 1 second bug reappeared?
> As the title states there
)
> As the title states there is a very bad bug in the BOINC CC:
>
> Since the SETI server nearly constantly ignores requests for work at the
> moment my clients start to suck up units from Einstein (at least one Einstein
> unit for every unresponded request to SETI). The queue with Einstein units is
> now nearly too big for the completion time. When SETI comes back to life the
> client will suck SETI units and then won't be able to complete the Einstein
> units in time.
>
> This is a big bug in the concept! :(
> Can Einstein somehow slow down giving out work to clients, that aren't able to
> finish the work in time?
If the resource shares for SETI@Home and E@H are equal, then the scheduler should only send as much work as you can do in time t/2, where t is the time available for all BOINC work. But if the core client is not handing this resource share fraction correctly to the scheduler, then you'll get time t instead. Can you review your resource share fractions please?
Bruce
Director, Einstein@Home
> As the title states there
)
> As the title states there is a very bad bug in the BOINC CC:
>
> Since the SETI server nearly constantly ignores requests for work at the
> moment my clients start to suck up units from Einstein (at least one Einstein
> unit for every unresponded request to SETI). The queue with Einstein units is
> now nearly too big for the completion time. When SETI comes back to life the
> client will suck SETI units and then won't be able to complete the Einstein
> units in time.
>
> This is a big bug in the concept! :(
> Can Einstein somehow slow down giving out work to clients, that aren't able to
> finish the work in time?
If in one project the scheduler is not responding it is a good idea NOT to set the resource share on that server but on the server where you can make communication.
Greetings from Belgium
Thierry
Sorry, had some work to
)
Sorry, had some work to do...
I have CC 4.19 and share is set to 50/50 Seti/Einstein and connect every 3 days (meanwhile reduced to 1 day). After doing some math i think the problem lies herein:
The BOINC CC estimates ~12.5 hrs per einstein unit where in reality it uses ~16 hrs on my 1000 Mhz Pentium III-M notebook.
Aloha, Uli
> Sorry, had some work to
)
> Sorry, had some work to do...
>
> I have CC 4.19 and share is set to 50/50 Seti/Einstein and connect every 3
> days (meanwhile reduced to 1 day). After doing some math i think the problem
> lies herein:
> The BOINC CC estimates ~12.5 hrs per einstein unit where in reality it uses
> ~16 hrs on my 1000 Mhz Pentium III-M notebook.
Then reducing the cache to 2.34 days (3*12.5/16) should help adjust for this.
Cheers,
Bruce
Director, Einstein@Home