This was starting with an invalid memory alloc request size, like in
<https://www.postgresql.org/message-id/2080966.1592844878%40sss.pgh.pa.us>.
I followed the suggestion to trace down the faulty record, found and
fixed it. Now I can access that record again, but if I try to dump
the table I get:
| 2022-11-06 11:52:36.367 CET [2098-35] LOG: server process (PID 2964738) was terminated by signal 11: Segmentation
fault
| [...]
| 2022-11-06 11:53:46.229 CET [2964744-2] LOG: database system was not properly shut down; automatic recovery in
progress
| 2022-11-06 11:53:46.263 CET [2964744-3] LOG: redo starts at 7D/E53B6440
| 2022-11-06 11:53:46.265 CET [2964744-4] LOG: invalid record length at 7D/E5412958: wanted 24, got 0
| 2022-11-06 11:53:46.265 CET [2964744-5] LOG: redo done at 7D/E5412920
| 2022-11-06 11:53:46.749 CET [2098-39] LOG: database system is ready to accept connections
Whatever was "done" does not prevent the problem from being
reproducible. I tried to track down this faulty record as well, but
binary proves difficult with 25*10^6 rows if the server crashes and
rebuilds after each try.
Is there any reasonable way to proceed from here?
The system seems to be working fine (the currupted table contains
textual information only), but right now I cannot even make proper
backups which is not a good thing.
Regards,
Stefan