Re: PROBLEM Service Alert: hostname/check_postgres_old_transaction is CRITICAL **

On 2014-9-19 20:33 , Adarsh Sharma wrote:
> It returns approx *311 MB* data to the client servers.
>
> root > netstat -p | grep 45355
> tcp 0 1531648 localhost:5499 localhost:48848 ESTABLISHED 45355/postgres
>
> root > strace -p 45355
> -- Remain stuck for *2-3 *hours
> read(306, "\30\0kB\3\0\0\0\0\0\377\5\0\0\26\0\0\0\2\0\0\0\30\0+\264\3\0\0\0\0\0"..., 8192) = 8192
> sendto(10, "4174\0\0\0\0041535D\0\0\0\30\0\2\0\0\0\006248710\0\0\0"..., 8192, 0, NULL, 0
>
> Then after 2-3 hours it got completed automatically with below stacktrace :
>
> # strace -p 45355
> Process 45355 attached - interrupt to quit
> sendto(10, "4174\0\0\0\0041535D\0\0\0\30\0\2\0\0\0\006248710\0\0\0"..., 8192, 0, NULL, 0
>
> ) = -1 ECONNRESET (Connection reset by peer)

I'd guess that the receiving program crashes. It might not be able to stomach the 311 MB that's being sent to it, maybe it's expanding, and swapping, until it's killed by the OOM killer?

Doesn't look like a postgres problem to me. postgres is stuck trying to send data... try stracing the client to see what it does? (use "lsof -i :48848" or whatever the port number of the remote is to find the pid).

--
Jan-Pieter Cornet <johnpc(at)xs4all(dot)net>
"Any sufficiently advanced incompetence is indistinguishable from malice."
- Grey's Law