For anybody not 100% crystal clear about the meaning of 'Validate error' please read the opening post in this thread.
In the case Alex has linked to, it seems likely that the problem will be 'bad data'. The validator will not have looked at the third result yet. It will do so when the fourth is returned and the chances are that there will probably be a 'sanity check' failure at that point.
If there is, it's almost certain to be a problem with the data being crunched rather than any problem with the hosts themselves. I have seen previous examples of quorums containing two validate errors but once it gets to three, it's very unlikely to be a host problem.
For anybody not 100% crystal clear about the meaning of 'Validate error' please read the opening post in this thread.
In the case Alex has linked to, it seems likely that the problem will be 'bad data'. The validator will not have looked at the third result yet. It will do so when the fourth is returned and the chances are that there will probably be a 'sanity check' failure at that point.
If there is, it's almost certain to be a problem with the data being crunched rather than any problem with the hosts themselves. I have seen previous examples of quorums containing two validate errors but once it gets to three, it's very unlikely to be a host problem.
I've reported all three linked examples of multiple validate errors to the Devs. If you happen to see any more quorums like these, send a link in a PM to Bernd/HB/Oliver to draw attention. It's possible that all three have come from the same data set but I don't know how to tell that from the task names. In any case, if a data set is bad (contaminated with RFI?) there are bound to be more examples. When this has happened previously, all outstanding tasks that are affected get cancelled by the server.
It is not uncommon that we have a few bad BRP4 "beams" every month that slip through pre-processing without being detected as such. Most of the tasks generated from these will end up as validate errors.
Normally we do have scripts and internal web pages that monitor these, and it's usually me who then cancels the respective workunits.
However both monitoring and canceling currently relies strongly on our "web replica", a database that continues to occasionally slow down to an extent that makes it barely usable for certain operations. You can see this e.g. in the difference between the time that is shown on top of the server status page and the time you actually see it. This delay reflects the run time of the queries on this database.
Thus currently neither monitoring nor cancelling such beams works as it should. I'll have to change a thing or two on the web code to make this working again, possibly using the master DB or another replica, optimize the queries or whatever. But I can only do this when I'm back on my desk on Monday.
Recently, I`ve got 3 tasks completed but marked as `invalid`.
As far as I can remember they are from `Binary Radio Pulsar Search (Arecibo) v1.33 (BRP4X64)` category, being Task id #38571155 the most recent one.
I have a computer running, almost full time, without any significant interruptions.
Op sys is Ubuntu 13.04 64-bit, updated, with a ATI HD6450 GPU, also updated
with the latest AMD drivers. BOINC client/manager is 7.0.65 (available from Ubuntu
repositories).
Question: why am I getting those status of `Completed, but marked as invalid`?
Unfortunately it looks like that task 38571155 has been retired from the database so it isn't possible to comment on that particular task.
"Marked as invalid" just means it didn't pass some tests, and it seems like everyone will has some failures (I've got 6 failures at the moment).
Lots of things can cause the occasional failure (software problems with BOINC, unusual data in a workunit, hardware problems at the client, set-up problems on the server side). If it happens a lot, post a link to some failed work as soon as you can and hopefully someone here can help.
THX for the
)
THX for the answers!
Alexander
For anybody not 100% crystal
)
For anybody not 100% crystal clear about the meaning of 'Validate error' please read the opening post in this thread.
In the case Alex has linked to, it seems likely that the problem will be 'bad data'. The validator will not have looked at the third result yet. It will do so when the fourth is returned and the chances are that there will probably be a 'sanity check' failure at that point.
If there is, it's almost certain to be a problem with the data being crunched rather than any problem with the hosts themselves. I have seen previous examples of quorums containing two validate errors but once it gets to three, it's very unlikely to be a host problem.
Cheers,
Gary.
@ admins maybe you should
)
@ admins
maybe you should take a look onto this wu:
http://einsteinathome.org/workunit/144505455
seems to be one of these 'uncunchables'.
RE: For anybody not 100%
)
So with 4 errors 1 waiting and 1 not done yet... http://einsteinathome.org/workunit/144505444
Bad data or will know after the other 2 gets done... I linked this afew post back but didn't use the url tag...
PC setup MSI-970A-G46 AMD FX-8350 8 core OC'd 4.45GHz 16GB ram PC3-10700 Geforce GTX 650Ti Windows 7 x64 Einstein@Home
I've reported all three
)
I've reported all three linked examples of multiple validate errors to the Devs. If you happen to see any more quorums like these, send a link in a PM to Bernd/HB/Oliver to draw attention. It's possible that all three have come from the same data set but I don't know how to tell that from the task names. In any case, if a data set is bad (contaminated with RFI?) there are bound to be more examples. When this has happened previously, all outstanding tasks that are affected get cancelled by the server.
Cheers,
Gary.
I found 3 more of these
)
I found 3 more of these uncrunchables.
http://einsteinathome.org/workunit/144507334
http://einsteinathome.org/workunit/144377137
http://einsteinathome.org/workunit/144507599
All three pm'ed to BM.
Maybe BM is not online, so I
)
Maybe BM is not online, so I repost it and PM it also to Bikeman.
One of them has reached a level where some kind of selfdestruction should take place ...
https://dl.dropbox.com/u/50246791/Einstein%20bad%20wu1.PNG
http://einsteinathome.org/workunit/144507334
http://einsteinathome.org/workunit/144377137
http://einsteinathome.org/workunit/144507599
It is not uncommon that we
)
It is not uncommon that we have a few bad BRP4 "beams" every month that slip through pre-processing without being detected as such. Most of the tasks generated from these will end up as validate errors.
Normally we do have scripts and internal web pages that monitor these, and it's usually me who then cancels the respective workunits.
However both monitoring and canceling currently relies strongly on our "web replica", a database that continues to occasionally slow down to an extent that makes it barely usable for certain operations. You can see this e.g. in the difference between the time that is shown on top of the server status page and the time you actually see it. This delay reflects the run time of the queries on this database.
Thus currently neither monitoring nor cancelling such beams works as it should. I'll have to change a thing or two on the web code to make this working again, possibly using the master DB or another replica, optimize the queries or whatever. But I can only do this when I'm back on my desk on Monday.
BM
BM
Recently, I`ve got 3 tasks
)
Recently, I`ve got 3 tasks completed but marked as `invalid`.
As far as I can remember they are from `Binary Radio Pulsar Search (Arecibo) v1.33 (BRP4X64)` category, being Task id #38571155 the most recent one.
I have a computer running, almost full time, without any significant interruptions.
Op sys is Ubuntu 13.04 64-bit, updated, with a ATI HD6450 GPU, also updated
with the latest AMD drivers. BOINC client/manager is 7.0.65 (available from Ubuntu
repositories).
Question: why am I getting those status of `Completed, but marked as invalid`?
Thank you.
Free-DC statistics
Unfortunately it looks like
)
Unfortunately it looks like that task 38571155 has been retired from the database so it isn't possible to comment on that particular task.
"Marked as invalid" just means it didn't pass some tests, and it seems like everyone will has some failures (I've got 6 failures at the moment).
Lots of things can cause the occasional failure (software problems with BOINC, unusual data in a workunit, hardware problems at the client, set-up problems on the server side). If it happens a lot, post a link to some failed work as soon as you can and hopefully someone here can help.