Thread: pg_dump core dump, upgrading from 6.5b1 to 5/24 snapshot
Hi, I posted a related question yesterday to the general list, but Bruce Momjian <maillist@candle.pha.pa.us> suggested it belongs on the hackers list. Using pg_dump with 6.5b1 on solaris sparc crashes with a core dump. This means I can't keep backups and I can't upgrade my data model without being able to export the old data. On Bruce's suggestion I tried upgrading to a recent snapshot (in this case the one from 5/24). When I killed the old postmaster and started the new postmaster I got the following error in psql: bboard=> \d ERROR: nodeRead: Bad type 0 which sounds to me like maybe there was a change in the data format. Since I can't use pg_dump to get my data out of 6.5b1 I'm a bit stuck now. For now I've (quickly) gone back to using 6.5b1 and not the snapshot. If one of the developers wants debug info on the coredump let me know what you need (e.g., what commands to do in gdb--though I'll have to install this or get run permissions from the sysadmin). I did a quick stack trace in adb and it was dead in a call to fprintf, which I know does nothing to help pinpoint the bug but it sounds like a typical way to kill a C program. -- Ari Halberstadt mailto:ari@shore.net <http://www.magiccookie.com/> PGP public key available at <http://www.magiccookie.com/pgpkey.txt>
Ari Halberstadt <ari@shore.net> writes:
> Using pg_dump with 6.5b1 on solaris sparc crashes with a core dump. This
> means I can't keep backups and I can't upgrade my data model without being
> able to export the old data.
> On Bruce's suggestion I tried upgrading to a recent snapshot (in this case
> the one from 5/24). When I killed the old postmaster and started the new
> postmaster I got the following error in psql:
Hi Ari, I believe there's been at least one initdb-forcing change since 6.5b1,
so you're right, you can't run a newer postmaster until you have
extracted your data.
But I think what Bruce really had in mind was to run pg_dump out of
the latest snapshot against your 6.5b1 postmaster. That should work
as far as compatibility issues go. If we're really lucky, the bug has
been fixed since then --- if not, it'd be easiest to try to find it in
current sources anyway...
The coredump is happening in pg_dump, right, not in the connected
backend?
regards, tom lane
Tom Lane <tgl@sss.pgh.pa.us> wrote: >... > But I think what Bruce really had in mind was to run pg_dump out of >the latest snapshot against your 6.5b1 postmaster. That should work Running the new pg_dump: $ pg_dump -p 4001 -D -f bboard.out bboard Segmentation Fault (core dumped) > The coredump is happening in pg_dump, right, not in the connected >backend? Yes, it's pg_dump, not the backend. -- Ari Halberstadt mailto:ari@shore.net <http://www.magiccookie.com/> PGP public key available at <http://www.magiccookie.com/pgpkey.txt>
Ari Halberstadt <ari@shore.net> writes:
> Running the new pg_dump:
> $ pg_dump -p 4001 -D -f bboard.out bboard
> Segmentation Fault (core dumped)
>> The coredump is happening in pg_dump, right, not in the connected
>> backend?
> Yes, it's pg_dump, not the backend.
OK, could you see whether omitting -D and/or -f makes a difference?
regards, tom lane