The directory we got problems with is the "../archives" directory. 
Apparently it happened during a server move when the permissions weren't 
transferred correctly. I'll start looking into this again when I fixed 
some more pressing problems ...

Bernd

James Wanless wrote, On 21.04.2009 10:04 Uhr:
> [pls bear in mind I don't know how 'standard' my setup is, but I tried 
> this on my dummy project, with similar results for the old (but fairly 
> recent) server_stable, compared to the developmental changeset version 
> 17851. Also, pls bear in mind I have not used the -gzip flag with 
> db_purge before...]
> Anyway,
>
> current server_stable:
> db_purge -min_age_days 7:
> 209wu/464res archived, files in ../archives are of sizes, 
> 966307,8708,310859,3863 ie all present
> db_purge -min_age_days 7 -gzip:
> 209wu/464res archived, files in ../archives are of sizes, 98304,0,0,0 
> ie exhibiting the bug under consideration, presumably
>
> changeset 17851:
> db_purge -min_age_days 7:
> 209wu/464res archived, files in ../archives are of sizes, 
> 966307,8708,310859,3863 ie all present
> db_purge -min_age_days 7 -gzip:
> 209wu/464res archived, files in ../archives are of sizes, 98304,0,0,0 
> ie _still_ showing same bug present, apparently??
>
> What exactly is the directory that needs write access that Bernd 
> referred to - I can try changing permissions on that if appropriate?
> Also, in (all?) other respects, changeset 17851 checked out fine (at 
> least I didn't have any probelms w/ the normal routine of upgrade), 
> though I haven't done extensive testing, since this bug appears to be 
> the main issue for now...
> J
>
> On 21 Apr 2009, at 04:27, David Anderson wrote:
>
>> I checked in a fix to trunk.
>> Can someone please test this (including the unwriteable-dir case)
>> before I check it in to server_stable?
>> -- David
>>
>> Bernd Machenschalk wrote:
>>> Hi!
>>>
>>> If the current db_purge program is set to gzip the archives and
>>> encounters an error writing them (in our case the directory wasn't
>>> writable), it just writes the error in the log and continues to
>>> "archive" the results into nirvana:
>>>
>>> 2009-04-19 08:46:26.6223 [debug   ] Archived result [122476313] to a 
>>> file
>>> 2009-04-19 08:46:26.7304 [debug   ] Purged result [122476313] from 
>>> database
>>> 2009-04-19 08:46:26.7304 [debug   ] Archived workunit [49157937] to 
>>> a file
>>> 2009-04-19 08:46:26.8304 [debug   ] Purged workunit [49157937] from 
>>> database
>>> 2009-04-19 08:46:26.8305 [normal  ] Closed archive file
>>> ../archives/wu_archive_1240126442.xml.gz containing records of 1000
>>> workunits
>>> 2009-04-19 08:46:26.8305 [normal  ] Closed archive file
>>> ../archives/result_archive_1240126442.xml.gz containing records of 1000
>>> workunits
>>> 2009-04-19 08:46:26.8305 [normal  ] Closed archive file
>>> ../archives/result_index_1240126442.xml.gz containing records of 1000
>>> workunits
>>> 2009-04-19 08:46:26.8306 [normal  ] Closed archive file
>>> ../archives/wu_index_1240126442.xml.gz containing records of 1000 
>>> workunits
>>> 2009-04-19 08:46:26.8306 [normal  ] Closed archive files with 1000 
>>> workunits
>>> 2009-04-19 08:46:26.8310 [normal  ] Archived 1000 workunits and 4846 
>>> results
>>> 2009-04-19 08:46:29.9680 [normal  ] Opening archive
>>> ../archives/wu_archive_1240130789.xml.gz
>>> 2009-04-19 08:46:29.9687 [normal  ] Opening archive
>>> ../archives/result_archive_1240130789.xml.gz
>>> 2009-04-19 08:46:29.9693 [normal  ] Opening archive
>>> ../archives/result_index_1240130789.xml.gz
>>> sh: ../archives/wu_archive_1240130789.xml.gz: Permission denied
>>> 2009-04-19 08:46:29.9723 [normal  ] Opening archive
>>> ../archives/wu_index_1240130789.xml.gz
>>> sh: ../archives/result_archive_1240130789.xml.gz: Permission denied
>>> sh: ../archives/result_index_1240130789.xml.gz: Permission denied
>>> sh: ../archives/wu_index_1240130789.xml.gz: Permission denied
>>> 2009-04-19 08:46:29.9875 [debug   ] Archived result [119291362] to a 
>>> file
>>> 2009-04-19 08:46:30.2184 [debug   ] Purged result [119291362] from 
>>> database
>>> 2009-04-19 08:46:30.2185 [debug   ] Archived result [119291363] to a 
>>> file
>>> 2009-04-19 08:46:30.3623 [debug   ] Purged result [119291363] from 
>>> database
>>> 2009-04-19 08:46:30.3626 [debug   ] Archived result [120931871] to a 
>>> file
>>> 2009-04-19 08:46:30.4623 [debug   ] Purged result [120931871] from 
>>> database
>>> 2009-04-19 08:46:30.4625 [debug   ] Archived result [122480720] to a 
>>> file
>>>
>>> I think this is a bug, IMHO it should exit on such an error and stop
>>> purging the DB.
>>>
>>> However, I think here the error happens asynchronously, i.e. not on
>>> opening the pipe, as the command (gzip) itself apparently succeeds.
>>>
>>> The return values of the fprintf()s in archive_wu() shoud definitely be
>>> checked (and maybe there are some more places I haven't caught at first
>>> glance).
>>>
>>> If I am right, this fix should also go into the server_stable branch.
>>>
>>> Best,
>>> Bernd
>>>
>>> _______________________________________________
>>> boinc_dev mailing list
>>> [email protected]
>>> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
>>> To unsubscribe, visit the above URL and
>>> (near bottom of page) enter your email address.
>>
>> _______________________________________________
>> boinc_dev mailing list
>> [email protected]
>> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
>> To unsubscribe, visit the above URL and
>> (near bottom of page) enter your email address.
>
>

_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to