[Citadel Development] Re: Peak memory usage

2008-10-08 Thread davew
> > > > >But for my needs it comes down to can we just destoy a thread when a session >ends or do I need to have each session-specific allocation come from a new >mmap? > >Sam > > Oooh, one I can answer... In the server threads are created and destroyed based on load and stuff. In webcit I

[Citadel Development] Re: Peak memory usage

2008-10-06 Thread samjam
>Mon Oct 06 2008 07:28:38 EDT from [EMAIL PROTECTED] >Subject: Re: Peak memory usage > > how did you produce the load to get to the limit? jmeter? I haven't reached my memory limit, I fear to, I don't want a corrupted message store, or lost messages. So I restart citadel every few days when

[Citadel Development] Re: Peak memory usage

2008-10-06 Thread dothebart
how did you produce the load to get to the limit? jmeter?

[Citadel Development] Re: Peak memory usage

2008-10-05 Thread samjam
>Sat Oct 04 2008 22:20:29 EDT from IGnatius T [EMAIL PROTECTED] >Subject: Re: Peak memory usage > > >>1. alter citadel to try and use file buffers more (horrid job) >> > > Nope. We put a lot of energy into eliminating temp files wherever >possible. > > Couldn't you just swap to a page

[Citadel Development] Re: Peak memory usage

2008-10-04 Thread IGnatius T Foobar
>1. alter citadel to try and use file buffers more (horrid job) Nope. We put a lot of energy into eliminating temp files wherever possible. Couldn't you just swap to a page file, if your vhoster didn't give you a swap partition?

[Citadel Development] Re: Peak memory usage

2008-10-04 Thread samjam
>Sat Oct 04 2008 11:10:39 EDT from samjam >Subject: Re: Peak memory usage > > > >>Sat Oct 04 2008 08:47:08 EDT from [EMAIL PROTECTED] >>Subject: Re: Peak memory usage >> >> >> >>On XSLT; Maybe its apropriate for displaying uncertain HTML messages; for >>the rest of the content i'd say we've got

[Citadel Development] Re: Peak memory usage

2008-10-04 Thread samjam
>Sat Oct 04 2008 08:47:08 EDT from [EMAIL PROTECTED] >Subject: Re: Peak memory usage > > > >On XSLT; Maybe its apropriate for displaying uncertain HTML messages; for >the rest of the content i'd say we've got templates + css available now to >change the look & feel / behaviour of webcit; One insid

[Citadel Development] Re: Peak memory usage

2008-10-04 Thread dothebart
>Fr Okt 03 2008 16:14:35 EDT von [EMAIL PROTECTED] >Betreff: Re: Peak memory usage > > > >>Fri Oct 03 2008 05:00:41 EDT from [EMAIL PROTECTED] >>Subject: Re: Peak memory usage >> >> >> >>webcit has the concept of worker threads, much like apache. it increases >>/ decreasess the pool on need.

[Citadel Development] Re: Peak memory usage

2008-10-03 Thread samjam
>Fri Oct 03 2008 05:00:41 EDT from [EMAIL PROTECTED] >Subject: Re: Peak memory usage > > > >webcit has the concept of worker threads, much like apache. it increases >/ decreasess the pool on need. > >when a request comes in, the context_loop() parses it, and either goes to >logon and creates a

[Citadel Development] Re: Peak memory usage

2008-10-03 Thread dothebart
webcit has the concept of worker threads, much like apache. it increases / decreasess the pool on need. when a request comes in, the context_loop() parses it, and either goes to logon and creates a new wcsessios struct and puts it to the static list. The list is kept in global context, and is pro

[Citadel Development] Re: Peak memory usage

2008-10-03 Thread samjam
>Thu Oct 02 2008 20:30:35 EDT from [EMAIL PROTECTED] >Subject: Re: Peak memory usage > > > >> >> >>2. and - maybe this will work- hope that memory mapped files aren't locked >>into ram, and so maybe I could wrangle citadel to allocate memory using mmap >>instead of brk and that might side-step t

[Citadel Development] Re: Peak memory usage

2008-10-02 Thread dothebart
> > >2. and - maybe this will work- hope that memory mapped files aren't locked >into ram, and so maybe I could wrangle citadel to allocate memory using mmap >instead of brk and that might side-step the no-swap issue. > >Any advice for me? > >Sam i'd say if you find a good way to make strbuf u