[Linux-cluster] Ls and globbing taking ridiculously long on GFS

[Dylan Vanderhoof]

I received the following from Wendy in response to the question:

hmm.... I'm not familiar with this version of GFS but it is ok ... will 
see what oprofile says. A good tutorial can be found at:

http://people.redhat.com/wcohen/

The following is a cheat sheet for you - if you have questions, let us 
know. 

0. Obtain a script called "opreport_module" in Will's people page; put 
it into /tmp and chmod +x.
1. opcontrol --init
2. opcontrol --setup --vmlinux=/usr/lib/debug/lib/modules/`uname
-r`/vmlinux
3. opcontrol --start
4. (run your ls here)
5. opcontrol --dump
6. opcontrol --shutdown
7. opreport --long-filenames (pipe this output to a file and email me)
8. /tmp/opreport_module /gfs (pipe this output to a file and email me).

-- Wendy

-----Original Message-----
From: bigendian at gmail.com [mailto:bigendian at gmail.com] 
Sent: Wednesday, December 13, 2006 8:23 AM
To: linux clustering
Subject: Re: [Linux-cluster] Ls and globbing taking ridiculously long on
GFS


I know this is an old thread, but I'm curious how to run oprofile
myself.  Any tips?

Thanks,
Tom



On 11/7/06, Dylan Vanderhoof < DylanV at semaphore.com> wrote:
Certainly.

dylanv at iscsi0 /var/www/netresponse/lib/NetResponse/Controller $ gfs_tool

version
gfs_tool 1.03.00 (built Oct 17 2006 15:10:45)
Copyright (C) Red Hat, Inc.  2004-2005  All rights reserved.

dylanv at iscsi0 /var/www/netresponse/lib/NetResponse/Controller $ uname -a
Linux iscsi0 2.6.17-gentoo-r8 #8 SMP Fri Sep 15 13:57:05 PDT 2006 i686
Intel(R) Xeon(TM) CPU 3.00GHz GNU/Linux

As long as its not too disruptive, I can definitely run oprofile.
Instructions would be handy however.  =)

-Dylan

> -----Original Message-----
> From: Wendy Cheng [mailto:wcheng at redhat.com]
> Sent: Tuesday, November 07, 2006 11:00 AM
> To: linux clustering
> Subject: Re: [Linux-cluster] Ls and globbing taking
> ridiculously long on GFS
>
>
> Dylan Vanderhoof wrote:
> > I understand that a df, or ls -l that requires statting
> files should be 
> > slow.  However, I'm seeing ridiculous performance of just an ls, or
> > anything doing file globbing in directory reads.
> >
> My guess is that you have lots of small writes before the "ls" that 
> generates the disk flushing. Could you pass your kernel and
> gfs versions
> ? Mind running oprofile on your node (I can pass the
> instructions if you
> like) so I can take a look ?
> 
> -- Wendy
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>

--
Linux-cluster mailing list
Linux-cluster at redhat.com
https://www.redhat.com/mailman/listinfo/linux-cluster