Vmware ESXi (3.5) and RHEL (5.1) : Timekeeping Woes

Roehrig, Jack (John) Jack.Roehrig at ask.com
Wed Sep 16 11:29:55 CDT 2009

As stated earlier, the check for swapping is pretty easy. Just use the esxtop command: usr/bin/esxtop -b -d 2 -n 1 | cut -d',' -f 40 | grep -v esx | tr -d '"' | awk '{printf "%.0f",$1}'

For more accurate detection, try monitoring both that ESX value and time skew value on your monitoring system. We always see a causal relationship in time frequency offset in PPM fluctuation to VM swap usage, but don't always see a correlation with time skew and swapping. 

As a side note, we see many VMs whose memory allocation is a couple gigs less than utilization. The ones that we see swapping are using around 1/5 to 1/20 of their allocated RAM. So a 2GB machine would use between 100-250MB and a 6GB machine would use between 1 and 1.5GB.

If anyone decides to monitor the swapping / time skew / clock frequency correlations in their environment, I'd love to see RRD data. I'm curious if other people are experiencing this problem.

-Jack Roehrig

-----Original Message-----
From: Dan Pritts [mailto:danno at internet2.edu] 
Sent: Wednesday, September 16, 2009 9:24 AM
To: Roehrig, Jack (John)
Cc: Brian O'Mahony; linux-poweredge at lists.us.dell.com
Subject: Re: Vmware ESXi (3.5) and RHEL (5.1) : Timekeeping Woes

On Wed, Sep 16, 2009 at 11:04:37AM -0400, Roehrig, Jack (John) wrote:
> Without knowing how the swapping algorithm works, it's difficult to be certain under what conditions the swap will be used. The conditions look correct though. I have seen many 2, 4, 6, and 8GB RHEL5.1 guests whose swap file is utilized by their hosts. 

I don't know either, but FWIW i have ESX 3.5 servers.  My guest memory
subscription is a couple gigs less than the total RAM on the server, and the guests
generally don't all use all their allocated RAM.

they don't use a SAN, local disk only, and have a raid1 of 15k disks
dedicated to the ESX OS and guest swapfiles.  I don't have virtualcenter,
so i only get 24 hours of performance data, but whenever i happen to look, 
I don't see lots of disk i/o on that partition.

guests are a mix of a few win2008, some rhel4 32bit, many rhel5 64bit.  

It's certainly possbile that the system is swapping but it doesn't seem
likely to me.

I've still seen plenty of clock synchronization issues.


More information about the Linux-PowerEdge mailing list