high reported load average on idle EC2 instances

There is an issue with load average on Lucid EC2 instances where load average gets high despite no CPU usage, disk I/O or other usual causes.

This has been reproduced on several of the official AMIs both 64 bit and 32 bit, EBS and not (personally I've tried ami-714ba518, ami-4b4ba522, ami-fd4aa494).

My initial method of demonstrating the behavior is as follows (from my initial report in [1]):

If I start up a fresh server based on the new 10.04 64-bit image (ami-4b4ba522) and instal postgres-8.4 I quickly see the load average go up to somewhere consistently between 0.7 and 1.2.

I've also tried a non-EBS version of the same image (ami-714ba518) and as soon as i install postgres the load goes up and doesn't come down. Just to be clear, I haven't even created a database on these servers, and running top shows no CPU utilization.

After stopping postgres load drops rather quickly (down to 0.05 or so like I'd expect). I've spun up a few instances to test on and each one has shown the same behavior.

This behavior has also been demonstrated on a server running mysql with no postgres [2]

Two related threads on the ec2-ubuntu google group:
[1] http://groups.google.com/group/ec2ubuntu/browse_thread/thread/a7e9bc45cf923f8c
[2] http://groups.google.com/group/ec2ubuntu/browse_thread/thread/4be26e81b7c597bc

ProblemType: Bug
DistroRelease: Ubuntu 10.04
Package: linux-ec2 2.6.32.305.6
ProcVersionSignature: Ubuntu 2.6.32-305.9-ec2 2.6.32.11+drm33.2
Uname: Linux 2.6.32-305-ec2 x86_64
Architecture: amd64
Date: Tue May 4 14:22:52 2010
Ec2AMI: ami-4b4ba522
Ec2AMIManifest: (unknown)
Ec2AvailabilityZone: us-east-1a
Ec2InstanceType: m1.large
Ec2Kernel: aki-0b4aa462
Ec2Ramdisk: unavailable
ProcEnviron:
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: linux-meta-ec2