Date: Thu, 19 Jan 2006 11:16:58 +0000 From: "M. L." <user@celeritystorm.com> To: freebsd-questions@FreeBSD.org Subject: Apache & PV entries Message-ID: <43CF752A.6080507@celeritystorm.com>
next in thread | raw e-mail | index | archive | help
Hello list,
I am configuring a very heavily used apache webserver, that required
some special needs.
This particular configuration needs to have at least 1024 httpds always
running. Reaching this number is not a problem, but whenever I stop
apache via apachectl stop, I notice all the httpds take a long time (1
min) to vanish - and meanwhile, top shows most of them locked.
The states I see in top are:
*vm ob
RUN
while the header shows:
1038 processes:332 running, 12 sleeping, 694 lock
The numbers vary a lot. Eventually, they get killed, but it takes too
much time.
After some searching, I found out the cause for this was running out of
PVs. I've added the following relevant options to the kernel:
options KVA_PAGES=768
options PMAP_SHPGPERPROC=5120
sysctl vm.zone | grep PV shows: (note: httpd stopped here)
(limit) (used) (free)
(requests)
PV ENTRY: 24, 84410445, 5522, 1632543, 5381679
So, if my assumptions are correct, and running out of PVs was the cause
(I did get that message about increasing PMAP_SHPGPERPROC before I
placed that in the kernel), there is a problem now. If you look at the
vm.zone output, there are 84,410,445 PVs that could exist, but the
system shows only 1,632,543 free.
This leads me to believe there is another thing to tune. I already have
a high PVs limit, but according to this output, it's not going further
than 1,6 millions.
My question is, what can I do to make the system use the full 84
millions, as I have not found any way to do this.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43CF752A.6080507>