> Hmm. If the messages are less than PIPE_BUF bytes long (4096 bytes on
> Linux) then the writes are supposed to be atomic. Can you
> check whether
> the interspersal cases involve messages whose total length (all lines)
> exceeds 4K?
Tom,
Some of them involve long messages (>4K), but there are many that do not
(like the ones I had posted at the start of this thread).
I am not entirely sure where to start counting? I have been counting
from the last "LOG: statement: " string immediately preceding the
corruption and I have sometimes >4K, sometimes under. Also what is the
beginning of the message? (after the string "statement: "?) In any case
I have plenty that are way under 4K.
The good news is that most of these don't trip the query analysis tool I
am using (pgFouine). I get about 50-80 entries like this in the logs per
day and only maybe 1 in 300-400 trips the tool. Even the ones that don't
trip the tool probably cause problems and skew the stats a bit, but
that's, I hope, negligible. Still it would be good to get it to do the
right thing.
Thanks!
George
