[mdlug] ESX3 IO Problems

[Dan Pritts]

my glance at the output you sent made me think that it was generated
by a script.

Interestingly, I just moved a few ISO images to a virgin 3.0.2 2950
yesterday; scp maxed out at 3MB/sec.  later i scp'd them from the
same source to a different target host (redhat) and i got more like
9MB/sec (basically maxed out the 100Mb/sec ethernet), so it wasn't my
sending system that was the problem.

so, basically, i can replicate your issue.

let me know what vmware says but i think their answer is going to
be that this is expected behavior, and for good performance you
need to use a SAN.

On a semi-related note, do you have anything in place to monitor
the dell RAID cards on your ESX servers (eg, notification of disk
failures)?  There's no useful info in /proc/scsi.

It looks like maybe i can get this via IPMI and ipmi serial-over-lan.
Have you done this?

actually as i go back and look at this, the 2850 has useful info
in /proc/megaraid but the 2950 doesn't have info in
/proc/scsi/megaraid_sas.  Doh.



On Wed, Aug 08, 2007 at 02:51:39PM -0500, Morris, Tim wrote:
> I don't have a script yet, I was just dd'ing up a bunch of different
> size files and testing the transfer times (date && cp filea filea.00 &&
> date, or time cp filea filea.00).  
> 
> 
> --  Tim Morris - UGS PLM Software
> 
> -----Original Message-----
> From: mdlug-bounces at mdlug.org [mailto:mdlug-bounces at mdlug.org] On Behalf
> Of Dan Pritts
> Sent: Wednesday, August 08, 2007 11:12 AM
> To: MDLUG's Main discussion list
> Subject: Re: [mdlug] ESX3 IO Problems
> 
> Agreed - that really sucks.  It might be within the realm of what I saw
> happenning as i transferred vm's between systems.
> 
> If you send me your script I'll try it on lightly-loaded 2850 here.
> 
> On Tue, Aug 07, 2007 at 01:52:53PM -0500, Morris, Tim wrote:
> > The problems I'm seeing are on the service console. I don't have any