We have a setup with central NFS server (using 64 NFS daemons at the moment) and 10 servers that mounts data via NFS4. The NFS server is using XFS, Linux version 2.6.32-5-amd64 (Debian 2.6.32-30). It's a VM with gigabit interface and it delivers around 50-80 MBit in peak time. It's a quad core machine with 8 Gig of memory.
>From time to time the load literally explodes on the NFS server and eventually also on the webservers since they can't access the content on the NFS server.
I found this post via Google and you mentioned that you had a similar problem with PHP include path - care to enlighten what the problem was and how you solved it?
Also - if possible - could I have a look at your sysctl.conf file ?
On Feb 27, 2011, at 1:47 PM, Matthew Macdonald-Wallace wrote:
> On Sat, 2011-02-26 at 16:52 -0500, izrodix wrote:
>> Hi Guys,
>> we try to migrate our apache 2.2 mod_fcgid with php-cgi configuration
>> to nginx + php-fpm from ports on our freebsd 8.2 stable system.We hosted
>> round 200-300 sites.
>> The result is high load of the server, lots of system calls and
>> mysql-server 5.1.55 has doubling his threads.
>> Any suggestion ?
> What's the filesystem format underneath? We've seen some issues recently
> with NFS performance if the PHP include path is in the wrong order which
> drove the load up to over 200 on a 24-core server with 16G RAM.
> We also saw some issues with the maximum memory limits set
> in /etc/sysctl.conf which created "false" load in a similar range on the
> same box.
> nginx mailing list