Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> I suggested that bgwriter may be the culprit, mainly because the log
> lines were not preceded by the log_line_prefix as the other lines in the
> log. See an extract here: http://rafb.net/paste/results/awxFnY15.html
Hmm, what are the logging configuration parameters here exactly?
> Thomas also mentioned that after the error first appeared, all queries
> started failing with the same error message. That does not make any
> sense to me; but maybe it could have to do with a corrupt buffer in the
> buffer freelist, which every backend tried to write but failed.
I have an old note to myself that persistent write errors could "clog"
the bgwriter, because I was worried that after an error it would
stupidly try to write the same buffer again instead of trying to make
progress elsewhere. (CVS tip might be better about this, I'm not sure.)
A dirty buffer for a file that doesn't exist anymore would certainly
qualify as a persistent failure.
> I guess the important question to be asking is how did the system get
> into that state.
Yeah.
regards, tom lane
