Thank you Les, John, for the reply.
One thing that I noticed is that there are lots of nfs connections in TIME_WAIT. And since I have been facing problems with NFS (ie. taking too much time to cd to a mounting area on client) the problems regarding network appeared.
The TIME_WAIT connections does not free memory, rigth? Do you believe that NFS might be creating a network congestion? Among NFS optimization possibilities is that one in particular that take more effect to avoid performance problems in your oppinion?
Thanks,
Patricia
--- Em qui, 24/4/08, Les Mikesell <lesmikesell@gmail.com> escreveu:
De: Les Mikesell <lesmikesell@gmail.com>
Assunto: Re: [CentOS] DNS problem (on NAT configuration)
Para:
"CentOS mailing list" <centos@centos.org>
Cc: npaci-rocks-discussion@sdsc.edu
Data: Quinta-feira, 24 de Abril de 2008, 18:56
Patricia Bittencourt wrote:
>
> Hi,
>
> I'm dealing with a problem that the worker nodes that are
behind
> a NAT aren't able to reach outside from time to time. (ie: on a given
> moment I can ping an address name and immediately after I cannot:
"ping:
> unknown host").
> The NAT was configurated using the following:
> *nat
> -A POSTROUTING -o eth1 -j MASQUERADE
> COMMIT
>
> Please, any advice here would be appreciated.
>
The obvious fix is to configure your NAT gateway host as a caching DNS
server and make the hosts behind NAT use it's private address instead of
something on the other side. However, it should work anyway. Do you
have enough traffic that you could
be filling your conntrack tables? UDP
entries remain for a timeout interval since you can't see a disconnect.
See if the /proc/sys/net/ipv4/netfilter/ip_conntrack_* values are
reaonable for your scale.
--
Les Mikesell
lesmikesell@gmail.com
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos |