Thread: ftell error during pg_dump
Our IT administrator ran a pg_dump and received the following error: . . . pg_dump: dumping contents of table history pg_dump: [custom archiver] WARNING: ftell mismatch with expected position -- ftell used pg_dump: dumping contents of table history_archive pg_dump: [custom archiver] WARNING: ftell mismatch with expected position -- ftell used pg_dump: dumping contents of table historymetadata pg_dump: [custom archiver] WARNING: ftell mismatch with expected position -- ftell used . . . This did not happen on every table, but once it did happen it was on all subsequent tables. Any ideas?
William Garrison <postgres@mobydisk.com> writes:
> Our IT administrator ran a pg_dump and received the following error:
> pg_dump: [custom archiver] WARNING: ftell mismatch with expected
> position -- ftell used
What platform, and exactly what version of pg_dump? Is it possible
you ran out of disk space partway through?
regards, tom lane
I'm embarrassed to say it is 8.2.3 :( I'm not sure why they haven't upgraded our production servers to the latest 8.2 yet. It's running on Windows Server 2003, and it looks like there is plenty of disk space. I googled this and found someone reported defect 2461 for this, some time ago, but I'm not clear how to check the status of that bug. I subscribed to pgsql-bugs so I can ask there as well. http://archives.postgresql.org/pgsql-bugs/2006-06/msg00012.php I also checked the logs from yesterday and I don't see anything that was logged while the pg_dump was running. Tom Lane wrote: > William Garrison <postgres@mobydisk.com> writes: > >> Our IT administrator ran a pg_dump and received the following error: >> pg_dump: [custom archiver] WARNING: ftell mismatch with expected >> position -- ftell used >> > > What platform, and exactly what version of pg_dump? Is it possible > you ran out of disk space partway through? > > regards, tom lane > >
This is almost certainly the bug fixed in 8.2.4 and listed in the release notes as: "Allow pg_dump to do binary backups larger than two gigabytes on Windows (Magnus) " If it happens to be that your dump could approach the 2Gb limit, I suggest you upgrade to 8.2.9 and see if it goes away. As this is entirely a client bug, there would be nothing in the logs. //Magnus William Garrison wrote: > I'm embarrassed to say it is 8.2.3 :( I'm not sure why they haven't > upgraded our production servers to the latest 8.2 yet. It's running on > Windows Server 2003, and it looks like there is plenty of disk space. > > I googled this and found someone reported defect 2461 for this, some > time ago, but I'm not clear how to check the status of that bug. I > subscribed to pgsql-bugs so I can ask there as well. > http://archives.postgresql.org/pgsql-bugs/2006-06/msg00012.php > > I also checked the logs from yesterday and I don't see anything that was > logged while the pg_dump was running. > > Tom Lane wrote: >> William Garrison <postgres@mobydisk.com> writes: >> >>> Our IT administrator ran a pg_dump and received the following error: >>> pg_dump: [custom archiver] WARNING: ftell mismatch with expected >>> position -- ftell used >>> >> >> What platform, and exactly what version of pg_dump? Is it possible >> you ran out of disk space partway through? >> >> regards, tom lane >> >> >
yaaaaaaaaaaaaaaaaay! Thank you. Magnus Hagander wrote: > This is almost certainly the bug fixed in 8.2.4 and listed in the > release notes as: > "Allow pg_dump to do binary backups larger than two gigabytes on Windows > (Magnus) " > > If it happens to be that your dump could approach the 2Gb limit, I > suggest you upgrade to 8.2.9 and see if it goes away. > > > As this is entirely a client bug, there would be nothing in the logs. > > > //Magnus > > William Garrison wrote: > >> I'm embarrassed to say it is 8.2.3 :( I'm not sure why they haven't >> upgraded our production servers to the latest 8.2 yet. It's running on >> Windows Server 2003, and it looks like there is plenty of disk space. >> >> I googled this and found someone reported defect 2461 for this, some >> time ago, but I'm not clear how to check the status of that bug. I >> subscribed to pgsql-bugs so I can ask there as well. >> http://archives.postgresql.org/pgsql-bugs/2006-06/msg00012.php >> >> I also checked the logs from yesterday and I don't see anything that was >> logged while the pg_dump was running. >> >> Tom Lane wrote: >> >>> William Garrison <postgres@mobydisk.com> writes: >>> >>> >>>> Our IT administrator ran a pg_dump and received the following error: >>>> pg_dump: [custom archiver] WARNING: ftell mismatch with expected >>>> position -- ftell used >>>> >>>> >>> What platform, and exactly what version of pg_dump? Is it possible >>> you ran out of disk space partway through? >>> >>> regards, tom lane >>> >>> >>> >> >> > > >
William Garrison <postgres@mobydisk.com> writes:
> I'm embarrassed to say it is 8.2.3 :( I'm not sure why they haven't
> upgraded our production servers to the latest 8.2 yet. It's running on
> Windows Server 2003, and it looks like there is plenty of disk space.
Hmm. There was an 8.2.4 bug fix for pg_dump on Windows, but the log
message for it claimed it'd only affect dump files larger than 2GB,
and I'm not sure whether the symptoms matched this anyway. How big
is your dump exactly?
regards, tom lane
The dump is over 3GB. So there's no question this is it. I had a feeling this would all come down to not being on the latest version. Thanks to both Tom and Magnus for your help. Tom Lane wrote: > William Garrison <postgres@mobydisk.com> writes: > >> I'm embarrassed to say it is 8.2.3 :( I'm not sure why they haven't >> upgraded our production servers to the latest 8.2 yet. It's running on >> Windows Server 2003, and it looks like there is plenty of disk space. >> > > Hmm. There was an 8.2.4 bug fix for pg_dump on Windows, but the log > message for it claimed it'd only affect dump files larger than 2GB, > and I'm not sure whether the symptoms matched this anyway. How big > is your dump exactly? > > regards, tom lane > >