Re: 回复:Re: 回复:Re: [BUGS] BUG #5853: when the walsender cannot exit when reboot slave computer

(I repeat: please keep the mailing list cc'd so that other can help)

On 28.01.2011 08:49, zoulx1982 wrote:
> when the slave computer is reset , there are two situation:
> 1. the primary don't produce WAL, so walsender won't send any XLOG
> I use "netstat -anp | grep postgres" to find the connection state is ESTABLISHED
>
> 2. the primary produce WAL and need to send to walreceiver, but there need a long time to wait timeout(about 15 minutes)
> in this situation, the connection state is also ESTABLISHED

Yeah, 15 minutes is the timeout in TCP. I don't remember if that's just
a default that can be changed in the OS, or a requirement of the protocol.

> whether we should set a reasonable timeout to avoid waiting long time?

You can use tcp_keep_alive_* settings to somewhat alleviate that (see
manual
http://www.postgresql.org/docs/9.0/interactive/warm-standby.html#STREAMING-REPLICATION)
but other than that there's currently no application-level timeout.

A lingering walsender shouldn't normally cause any problems, though, it
will timeout eventually.

--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com