Hi to all,
those helped or enjoyedmy english. I waited over the day to give a summary.
> your pgsql backend must be able to accommodate the max persistent
> connections * the max number of apache daemons.
>
> the php pgsql.max_persistent setting is per apache daemon.
All of your answers where correct, i switched from "keep all persistant" to
"throw all away"
and the new hardware behaves very harmful. ;-)
Before this, on older hardware with apache on the same machine, it was a bad
idea not to use persistant links,
but with the "only-Db-Server" it runs very good.
I reconfigured my apache-machines back and now over the whole day i had load
< 2 , mostly < 1,2 , even although I was running several maintenance jobs on
the machine.
I guess, that I have had reached the limits for parallel postgres's (1.000)
on the hardware and the concept of so many databases (> 10) with a lot of
apache instances (> 100) running is not to be combined with connection
sharing.
Now i use following main options:
~~~~
vacuum_mem = 8192
sort_mem = 8192 # 8 MB
max_connections = 800 # even 20 will be enough
shared_buffers = 8192 # 64mb , for 1 GB RAM, total sum for all processes
~~~~
kernel.shmall = 1342177280
kernel.shmmax = 1342177280
~~~~
Thanx for hints,
Andre
