CentOS - Oct 2007 - Perfomance tuning for NIS client

Hi,

I have both linux and solaris NIS client against a Solaris NIS Server, but
my linux box are shamefully slow compare to solaris some sample:

on Solaris 5.8:

# ypwhich
transporter02.domain.com
# time id userid
uid=36923(userid) gid=36923(u_036923)

real        0.0
user        0.0
sys         0.0

on Linux:

[root at linux]# ypwhich
transporter02.domain.com
[root at amsdc2-n-s04taw root]# time id userid
uid=36923(userid) gid=36923(u_036923) groups=36923(u_036923)

real    0m10.509s
user    0m0.030s
sys     0m0.090s

********** Route****************

[root at linux]# traceroute transporter02
traceroute to transporter02.domain.com (IP), 30 hops max, 46 byte packets
 1  nlamsdc2mp078-hsrp.net-equip.domain.net (IP)  0.000 ms  0.000 ms  0.000ms
 2  transporter02.domain.com (IP)  0.000 ms  0.000 ms  0.000 ms

solaris# traceroute transporter02
traceroute to transporter02.domain.com (145.6.246.53), 30 hops max, 40 byte
packets
 1  transporter02.domain.com (145.6.246.53)  0.462 ms  0.279 ms  0.306 ms

Although I considered that latency involved but still I'm not statisfied, is
there any tunable parameter that I can set on linux box to somehow improve
the performance?

I also noticed things that I'm not sure if this is default on linux nis
bindings:

[root at linux]# pwd
/var/yp/binding
[root at linux]# ls
NONAME.DOMAIN.COM.1  NONAME.DOMAIN.COM.2  ===> why there are two of this on
linux?

solaris# cd /var/yp/binding
solaris# ls
NONAME.DOMAIN.COM  xprt.tcp.3        xprt.ticlts.3     xprt.ticotsord.2
xprt.udp.2
xprt.tcp.1        xprt.ticlts.1     xprt.ticots.3     xprt.ticotsord.3
xprt.udp.3
xprt.tcp.2        xprt.ticlts.2     xprt.ticotsord.1  xprt.udp.1
ypbind.pid

Thanks in advance for any feedback.

Things I've tried to tweak:

* put NIS Server IP on /etc/yp.conf
* put NIS Server IP on /etc/hosts
no positive result.


joseph
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.centos.org/pipermail/centos/attachments/20071013/64a0c72b/attachment-0004.html>

On Sat, 2007-10-13 at 03:45 +0800, joseph blase wrote:
> Hi,
>  
> I have both linux and solaris NIS client against a Solaris NIS Server,
> but my linux box are shamefully slow compare to solaris some sample:
>  
> on Solaris 5.8:
>  
> # ypwhich
> transporter02.domain.com
> # time id userid
> uid=36923(userid) gid=36923(u_036923)
>  
> real        0.0
> user        0.0
> sys         0.0
>  
> on Linux:
>  
> [root at linux]# ypwhich
> transporter02.domain.com
> [root at amsdc2-n-s04taw root]# time id userid
> uid=36923(userid) gid=36923(u_036923) groups=36923(u_036923)
>  
> real    0m10.509s
> user    0m0.030s
> sys     0m0.090s
> 
Do you have nscd running on the Linux box?  Enabling it will allow
caching of lookups to passwd, group and hosts.

-Steve

On Sat, Oct 13, 2007 at 03:45:14AM +0800, joseph blase
alleged:
> [root at amsdc2-n-s04taw root]# time id userid
> uid=36923(userid) gid=36923(u_036923) groups=36923(u_036923)
> 
> real    0m10.509s
> user    0m0.030s
> sys     0m0.090s
Start with 'strace id userid' to see where it is stalling.  Maybe
you'll find a
bad PTR lookup or something.

You can also determine if the stall is in NIS code with 'ypmatch userid
passwd'.

If nscd is running, does killing it make a difference?

Any selinux violations in /var/log/messages?


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<http://lists.centos.org/pipermail/centos/attachments/20071012/5330342c/attachment-0004.sig>