At 11:18 PM 18-11-2001 -0500, Tom Lane wrote:
>Lincoln Yeoh <lyeoh@pop.jaring.my> writes:
>> Trying to pg_dump the table gives me this:
>> <snipped>
>> ERROR: MemoryContextAlloc: invalid request size 1163153238
>> PQendcopy: resetting connection
>
>This looks like a corrupted-data problem ...
Yah. Timestamp was BC :).
>> I've truncated the table and it runs ok now.
>
>... but the evidence is now gone, so we can't really probe into it
>further :-(. You might be well advised to run some hardware diagnostics
>to see if you have any RAM problems, flaky disk controllers, that sort
>of thing. Not that Postgres has no bugs, of course, but we've seen
>quite a number of data-corruption reports that ultimately traced to
>hardware problems.
>The behavior during a SELECT seems odd also:
>
>> SELECT * from arch_ranks_arch4 ;
>> Backend message type 0x44 arrived while idle
>> pqReadData() -- backend closed the channel unexpectedly.
>
>This suggests that libpq and the backend got out of sync somehow,
>but I thought we'd fixed that class of problems years ago. If you
>can reproduce this it'd be worth looking into.
I'll try to upgrade to 7.1.3. Then if it happens again it'll mean more.
Cheerio,
Link.