Robert Haas escribió:
> A broader complaint I have with this patch is that it almost but
> not-quite solves a problem I've had a few times in the past: namely,
> searching through the data directory for data blocks which have LSNs
> in the future. This has come up a few times for me, and this tool
> would make it easier, because I'd be able to run it and look through
> the output to see which relations have high max-LSN values. However,
> it wouldn't be quite enough, because it'd only tell me about the block
> with the highest LSN in each file, whereas what I'd really want to
> find is every block with an LSN greater than some threshold value.
> Maybe I'm pushing the envelope too much by trying to fit that into the
> framework of this patch, but I can't help thinking we're not going to
> want both pg_computemaxlsn and pg_findlsnsaftersomethreshold that are
> 95% the same code, so maybe we ought to rename the utility to
> something slightly more generic than "pg_computemaxlsn".
Perhaps not coincidentally, I had a need to do this recently. Perhaps
we should turn the utility into a generic tool to report existing LSNs,
with options to 1) report only the highest one in a given file, 2)
report only those that exceed some threshold. So maybe pg_reportlsn or
pg_extractlsn.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
