[ec2] swap?

[Darren Govoni]

Hi Rob,
   According to the launchpad bug, it seems related to a bug in the
kernel at the moment, but as yet I am only assuming my issue is related
to the documented bug because of the description.

My new data shows this.

My system runs fine for about 15 minutes and I see the memory dwindle
down (I'm running 32 supervisord python app threads that make system
calls). For the most part, memory is fairly static. But inches down.

At one point it jumps from about 250MB free to 34MB free and that is the
last update I see after it freezes. My 'top' indicates that kswapd with
a nice of -5 has jumped to the top of the process list. Not sure if
thats because of the memory situation or the cause of it.

After that, the instance cannot be used, rebooted or contacted in
anyway.

I will post this on the launchpad thread too, which seems to indicate a
bug in the process scheduler.

But if kswapd is kicking in and there is no swap, I wonder what that
would mean.

Darren

On Mon, 2009-07-27 at 08:23 +1200, Robert Coup wrote:
> Hi Darren,
> 
> 
> On Mon, Jul 27, 2009 at 5:44 AM, Darren Govoni <darren at ontrenet.com>
> wrote:
>         
>         The problem I am running into may be related to the other
>         posted issue
>         here when instances freeze.
>         
>         https://bugs.launchpad.net/ubuntu/+source/linux/+bug/276476
>         
>         Basically, all of my instances will freeze up after some time
>         running
>         where they are reading/writing AWS services heavily as well as
>         local
>         disk. I was concerned if its related to no virtual memory
>         swap.
> 
> 
> If you have no swap things will get killed, be it apache, a database,
> or something else. The system won't go down if it runs out of memory.
> If the system hangs it implies (to me) that it's a filesystem or
> kernel problem. What filesystem are you running? Any RAID/LVM? Are you
> using EBS or is this the local instance storage?
> 
> 
>         
>         once it freezes I cannot even reboot it. Its gone forever.
> 
> 
> So rebooting via the AWS console/tools doesn't help? Note that it can
> easily take 5 minutes or longer to reboot an instance via the AWS
> restart command. After your instance has rebooted, there may be
> something informative (about the problem) in the EC2 console output.
> 
> 
> Rob :)
> 
>