Re: [ZFS] Server periodically become unavailable

On 20.10.2012 15:11, Vladislav Prodan wrote:
>
>> FreeBSD 9.1-PRERELEASE #0: Wed Jul 25 01:40:56 EEST 2012
>
> I have the server: 8 cores AMD, 16GB RAM, 4x3TB HDD in RAID10 for ZFS.
> Sometime wheels fall off the server and the network.
> Can this clean-up memory for ZFS cache?
> I enclose a picture with the monitoring system at the time lags.
>
> http://imageshack.us/a/img341/9643/memoryusage.png
> http://imageshack.us/a/img22/6935/nginxclientstat.png
> http://imageshack.us/a/img19/8817/realmemory.png
>
> #cat /etc/sysctl.conf
> kern.ipc.somaxconn=65535

Remove this. It doesn't do what you think it does.

> kern.ipc.maxsockets=204800
> net.inet.ip.portrange.first=1024
> net.inet.ip.portrange.last=65535
> kern.ipc.shmmax=67108864
> kern.ipc.shmall=67108864
> net.inet.tcp.rfc3465=0

Any particular reason why you turn this off?

> net.graph.maxdgram=8388608
> net.graph.recvspace=8388608
> net.route.netisr_maxqlen=4096
> kern.ipc.nmbclusters=400000
> kern.ipc.maxsockbuf=83886080

83MB for a socket buffer is too much unless you're doing some HPC stuff.
The default should be fine.

> net.inet.tcp.recvbuf_inc=524288

Again, this should be left at default unless you have a very specific
reason to increment it.

> net.inet.tcp.recvbuf_max=16777216

ditto.

> net.inet.tcp.sendbuf_inc=524288

ditto.

> net.inet.tcp.sendbuf_max=16777216

ditto.

> net.inet.tcp.sendspace=65536

This is the default setting.

> net.inet.tcp.keepidle=300000
> net.inet.tcp.keepintvl=60000
> net.inet.ip.fw.dyn_max=65535
> net.inet.ip.fw.dyn_buckets=65536
> net.inet.ip.fw.dyn_ack_lifetime=120
> net.inet.ip.fw.dyn_syn_lifetime=10
> net.inet.tcp.nolocaltimewait=1
> security.bsd.hardlink_check_uid=1
> security.bsd.hardlink_check_gid=1
> security.bsd.see_other_uids=0
> security.bsd.see_other_gids=0

-- 
Andre