Thread: error restoring large objects during pg_restore
I'm trying to pg_restore a database using a process that I've used a bunch (10-15) times in the past, and this is the first time it's ever failed on me. It's failing with the following error: [snip] pg_restore: creating table for large object cross-references pg_restore: [custom archiver] could not read data block - expected 1, got 0 pg_restore: *** aborted because of error When I look in the logs, I see the following: [snip] Jun 10 14:24:19 vault pgqv[16995]: [242519] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898427, 60983741); Jun 10 14:24:19 vault pgqv[16995]: [242520] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898503, 60983743); Jun 10 14:24:19 vault pgqv[16995]: [242521] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898534, 60983745); Jun 10 14:24:19 vault pgqv[16995]: [242522] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898544, 60983747); Jun 10 14:24:19 vault pgqv[16995]: [242523] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898552, 60983749); Jun 10 14:24:19 vault pgqv[16995]: [242524] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898564, 60983751); Jun 10 14:24:19 vault pgqv[16995]: [242525] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898570, 60983753); Jun 10 14:24:19 vault pgqv[16995]: [242526] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898584, 60983755); Jun 10 14:24:19 vault pgqv[16995]: [242527] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (57898592, 60983757); Jun 10 14:24:19 vault pgqv[16995]: [242528] DEBUG: query: Insert Into dump_blob_xref(oldOid, newOid) Values (1, 60983759); I've done this twice, with two different db dumps, and both have given me the same error. I'm wondering if that last line (with the oldOid value of 1) is significant. Thoughts, ideas? -ron
Ron Snyder <snyder@roguewave.com> writes:
> I'm trying to pg_restore a database using a process that I've used a bunch
> (10-15) times in the past, and this is the first time it's ever failed on
> me. It's failing with the following error:
> pg_restore: creating table for large object cross-references
> pg_restore: [custom archiver] could not read data block - expected 1, got 0
> pg_restore: *** aborted because of error
Hm. Looking at the code, this appears to be an unexpected-EOF kind of
error (it expected to be able to read 1 more byte of data, and there
wasn't any).
> Jun 10 14:24:19 vault pgqv[16995]: [242528] DEBUG: query: Insert Into
> dump_blob_xref(oldOid, newOid) Values (1, 60983759);
> I've done this twice, with two different db dumps, and both have given me
> the same error. I'm wondering if that last line (with the oldOid value of
> 1) is significant.
I think the "1" is coincidental. It does seem that there is something
broken about the format of the dump file though. What version pg_dump
were you using, exactly?
Other things to check would include accidental truncation or corruption
of the dump file (eg, copying it onto a Windows machine might result in
newline conversion, which would be bad).
Philip, any other ideas where to look?
regards, tom lane
This is with postgresql 7.2.1 (we're in the process of moving to 7.3.3). I believe that the chance of file corruption should be very low-- Windows boxes have no opportunity to touch this file. The server is a RH Linux box with logical volumes (both as storage for the dump files and the storage for the database). Thanks for looking! I'm currently "strace"ing the pg_restore to see if that lends any clues. -ron > -----Original Message----- > From: Tom Lane [mailto:tgl@sss.pgh.pa.us] > Sent: Tuesday, June 10, 2003 4:04 PM > To: Ron Snyder > Cc: 'pgsql-general@postgresql.org'; Philip Warner > Subject: Re: [GENERAL] error restoring large objects during > pg_restore > > > Ron Snyder <snyder@roguewave.com> writes: > > I'm trying to pg_restore a database using a process that > I've used a bunch > > (10-15) times in the past, and this is the first time it's > ever failed on > > me. It's failing with the following error: > > > pg_restore: creating table for large object cross-references > > pg_restore: [custom archiver] could not read data block - > expected 1, got 0 > > pg_restore: *** aborted because of error > > Hm. Looking at the code, this appears to be an unexpected-EOF kind of > error (it expected to be able to read 1 more byte of data, and there > wasn't any). > > > Jun 10 14:24:19 vault pgqv[16995]: [242528] DEBUG: query: > Insert Into > > dump_blob_xref(oldOid, newOid) Values (1, 60983759); > > > I've done this twice, with two different db dumps, and both > have given me > > the same error. I'm wondering if that last line (with the > oldOid value of > > 1) is significant. > > I think the "1" is coincidental. It does seem that there is something > broken about the format of the dump file though. What version pg_dump > were you using, exactly? > > Other things to check would include accidental truncation or > corruption > of the dump file (eg, copying it onto a Windows machine might > result in > newline conversion, which would be bad). > > Philip, any other ideas where to look? > > regards, tom lane >
Oh yeah, I forgot to mention that the partition the dumps are being written to are large enough to hold 5 times as much data as they've got. These dumps are the only thing of any consequence (size) that are being written to the partition, so the dumps aren't being truncated by out of disk space problems. -ron > -----Original Message----- > From: Ron Snyder > Sent: Tuesday, June 10, 2003 4:52 PM > To: 'Tom Lane'; Ron Snyder > Cc: 'pgsql-general@postgresql.org'; Philip Warner > Subject: RE: [GENERAL] error restoring large objects during > pg_restore > > > This is with postgresql 7.2.1 (we're in the process of moving > to 7.3.3). I believe that the chance of file corruption > should be very low-- Windows boxes have no opportunity to > touch this file. The server is a RH Linux box with logical > volumes (both as storage for the dump files and the storage > for the database). > > Thanks for looking! > > I'm currently "strace"ing the pg_restore to see if that lends > any clues. > > -ron > > > -----Original Message----- > > From: Tom Lane [mailto:tgl@sss.pgh.pa.us] > > Sent: Tuesday, June 10, 2003 4:04 PM > > To: Ron Snyder > > Cc: 'pgsql-general@postgresql.org'; Philip Warner > > Subject: Re: [GENERAL] error restoring large objects during > > pg_restore > > > > > > Ron Snyder <snyder@roguewave.com> writes: > > > I'm trying to pg_restore a database using a process that > > I've used a bunch > > > (10-15) times in the past, and this is the first time it's > > ever failed on > > > me. It's failing with the following error: > > > > > pg_restore: creating table for large object cross-references > > > pg_restore: [custom archiver] could not read data block - > > expected 1, got 0 > > > pg_restore: *** aborted because of error > > > > Hm. Looking at the code, this appears to be an > unexpected-EOF kind of > > error (it expected to be able to read 1 more byte of data, and there > > wasn't any). > > > > > Jun 10 14:24:19 vault pgqv[16995]: [242528] DEBUG: query: > > Insert Into > > > dump_blob_xref(oldOid, newOid) Values (1, 60983759); > > > > > I've done this twice, with two different db dumps, and both > > have given me > > > the same error. I'm wondering if that last line (with the > > oldOid value of > > > 1) is significant. > > > > I think the "1" is coincidental. It does seem that there > is something > > broken about the format of the dump file though. What > version pg_dump > > were you using, exactly? > > > > Other things to check would include accidental truncation or > > corruption > > of the dump file (eg, copying it onto a Windows machine might > > result in > > newline conversion, which would be bad). > > > > Philip, any other ideas where to look? > > > > regards, tom lane > > >