On 04.12.2010 01:30, Igor Galić wrote:
Hi Rainer,
this was an amazingly terse and most of all: Excellent description.
Can I use it in the Performance documentation and/or the mpm docs?
Of course!
Go for it and sorry about not having improved the docs myself already.
Regards,
Rainer
----- "Rainer Jung"<[email protected]> wrote:
On 26.11.2010 19:20, Jonas Eckerman wrote:
On 2010-11-25 15:28, Igor Galić wrote:
Worker MPM will *never* spawn more than StartServers children.
According to the docs at
<http://httpd.apache.org/docs/current/mod/mpm_common.html#startservers>
it will.
Quote:
---8<---
The StartServers directive sets the number of child server
processes
created on startup. As the number of processes is dynamically
controlled
depending on the load, there is usually little reason to adjust
this
parameter.
The default value differs from MPM to MPM. For worker the default
is
StartServers 3. For prefork defaults to 5 and for mpmt_os2 to 2.
---8<---
The *ServerLimit* directive oth puts a hard limit on the numkber of
children Apache can start.
Quote:
---8<---
For the worker MPM, this directive [ServerLimit] in combination
with
ThreadLimit sets the maximum configured value for MaxClients for
the
lifetime of the Apache process. Any attempts to change this
directive
during a restart will be ignored, but MaxClients can be modified
during
a restart.
---8<---
More info at:
<http://httpd.apache.org/docs/current/mod/mpm_common.html#serverlimit>
StartServers and the spare directives configure the dynamics. As their
name indicates, how many processes are started, how many threads
should
be idle minimum befoe sarting new processes and how many threads you
allow to idle before stopping processes.
The maximum number of connections that can be handled in parallel is
given by MaxClients. It is the most important parameter for sizing.
Then
ThreadsPerChild defines how powerful each process will be, or in other
words, over how many processes the MaxClient threads will be
distributed. If you increase it, you will need less processes and if
modules use e.g. caches per process those will be more efficient. On
the
other hand locking activity may increase and if you are using unstable
modules (crashes or leaks) more connections will break if one process
e.g. crashes.
The ServerLimit and ThreadLimit params only have to do with the
ability
of dynamic reconfiguration the above using only graceful restarts, not
full restarts. During graceful restart the main Apache process is not
recycled. It needs to maintain a memory table with columns and rows fr
processes and threads inside processes. The size of the table needs to
be big enough for any ThreadsPerChild and (MaxClients /
ThreadsPerChild)
you want to use.
So if no dynamic reconfiguration for those values is expected, you can
set ThreadLimit to ThreadsPerChild and ServerLimit to MaxClients /
ThreadsPerChild. If you plan to adjust ThreadsPerChild or MaxClients
to
higher values later (using a graceful restart), you need to plan ahead
and set ThreadLimit and ServerLimit to the max values you expect to
need
according to these formulas.
HTH.
Rainer
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
