Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

/var/lib/nova/instances RAM cache usage

I've been looking at RAM allocation on the compute hosts.

Using a tool, I see that In a number of environments, I have /var/lib/nova/instances using a lot of cache - to the tune of 119G (another env is using 135 GB). The files with the instance ID names are 49 GB total (3 files) and the files inside of the _base directory is using 68 G. So, between those three, 117 of 119 GB of cache are being used.

With all the cache use, the OS ends up using Swap. It seems that it doesn't want to remove the above data from cache.

In another environment,/var/lib/nova/instances is using only 628K of cache.

So, are we supposed to have so much of /var/lib/nova/instances cached? What is deciding to cache that? How can it be stopped / reduced?

We have id=drive-virtio-disk0,format=qcow2,cache=none set.

We're using Icehouse and can't change now.

Thanks!

/var/lib/nova/instances RAM cache usage

I've been looking at RAM allocation on the compute hosts.

Using a tool, I see that In a number of environments, I have /var/lib/nova/instances using a lot of cache - to the tune of 119G (another env is using 135 GB). The files with the instance ID names are 49 GB total (3 files) and the files inside of the _base directory is using 68 G. So, between those three, 117 of 119 GB of cache are being used.

With all the cache use, the OS ends up using Swap. It seems that it doesn't want to remove the above data from cache.

In another environment,/var/lib/nova/instances is using only 628K of cache.

So, are we supposed to have so much of /var/lib/nova/instances cached? What is deciding to cache that? How can it be stopped / reduced?

We have id=drive-virtio-disk0,format=qcow2,cache=none set.

We're using Icehouse and can't change now.

Thanks!

/var/lib/nova/instances RAM cache usage

I've been looking at RAM allocation on the compute hosts.

Using a tool, I see that In a number of environments, I have /var/lib/nova/instances using a lot of RAM cache - to the tune of 119G (another env is using 135 GB).

The files with the instance ID names /var/lib/nova/instances/'instanceID'/disk directories are 49 GB total (3 files) and the files). The files inside of the _base directory is using 68 G. So, between those three, 117 of 119 GB of cache are being used.

With all the cache use, the In this environment, I don't care if bringing up an instance takes somewhat longer. So, it would seem, removing _base from the RAM cache would only help me. I'm not sure how I could get there, though.

The OS ends up using Swap. It seems that it I'm wondering if the OS doesn't want to remove the above data from cache. If not the system using swap, I wouldn't care about the cache usage.

In another environment,/var/lib/nova/instances is using only 628K of cache.

So, are we supposed to have so much of /var/lib/nova/instances cached? What is deciding to cache that? How can it be stopped / reduced?

We have id=drive-virtio-disk0,format=qcow2,cache=none set.

We're using Icehouse and can't change now.

Thanks!

/var/lib/nova/instances RAM cache usage

I've been looking at RAM allocation on the compute hosts.

Using a tool, I see that In a number of environments, I have /var/lib/nova/instances using a lot of RAM cache - to the tune of 119G (another env is using 135 GB).

The /var/lib/nova/instances/'instanceID'/disk directories are 49 GB total (3 files). The files inside of the _base directory is using 68 G. So, between those three, 117 of 119 GB of cache are being used.

In this environment, Could _base be told not to sit in cache? Would just just impact VM startup (which I don't care if bringing up an instance takes somewhat longer. So, it would seem, removing _base from the RAM cache would only help me. about so much?) I'm not sure how I could get there, though.

The OS ends up using Swap. I'm wondering if the OS doesn't want to remove the above data from cache. If not the system using swap, I wouldn't care about the cache usage.

In another environment,/var/lib/nova/instances is using only 628K of cache.

So, are we supposed to have so much of /var/lib/nova/instances cached? What is deciding to cache that? How can it be stopped / reduced?

We have id=drive-virtio-disk0,format=qcow2,cache=none set.set, so why is /var/lib/nova/instances/'instanceID'/disk cached. Maybe a qemu bug in the version I'm running? I just posted to the qemu mailing list.

That doesn't account for the _base directory files being cached.

Recall, that neither of these locations are cached in another environment I have.. which is -supposed to be- configured in parallel.

We're using Icehouse and can't change now.

Thanks!