Re: Are indexes blown?

On Fri, Feb 15, 2008 at 8:36 AM, Phoenix Kiula <phoenix(dot)kiula(at)gmail(dot)com> wrote:

> No. They are the vmstat figures from when I was replying to your
> email. What will vmstat tell me and how should I set it up to do
> "vmstat 10 logging"?

Something like

vmstat 10 > vmstat.log

> LOG: could not receive data from client: Connection reset by peer
> LOG: unexpected EOF on client connection
> LOG: could not receive data from client: Connection reset by peer
> LOG: unexpected EOF on client connection
>
> Now I don't know what is wrong or even where I should look. Postgresql
> is often taking quite a bit of memory and CPU resources.
>
> I've reduced work_mem to 10MB and Max_connections to 100. (Anyway, the
> old values were working just fine until recently!)
>
> The biggest problem: when I su into postgres user and do a psql to get
> into the PG console in my SSH, it takes a whole lot of time to come
> up! It used to come up in a jiffy earlier!!! It now shows me this
> error:

How many pgsql processes are there when this happens? Try something like

ps axu|grep postgres

to see. use

ps axu|grep postgres|wc -l

to get a rough count. I'm guessing that your web service layer is
keeping old connections open. could be something as ugly as php's
pg_pconnect or a buggy jdbc driver, etc...