provisioning clients in teuthology with an extra local filesystem
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology: https://tracker.ceph.com/issues/6373 It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts. Is it possible to do this in teuthology? How would I approach this? Thanks, -- Jeff Layton <jlayton@redhat.com>
On 12/1/20 9:55 AM, Jeff Layton wrote:
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology:
https://tracker.ceph.com/issues/6373
It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts.
Is it possible to do this in teuthology? How would I approach this?
I think I can make this happen pretty easily in ceph-cm-ansible. What I'd need from you is desired filesystem type and size. Once I'm done with my end, we'll need to create a cephlab.ansible overrides yaml fragment to stick in that suite's qa directory.
On Tue, 2020-12-01 at 10:22 -0500, David Galloway wrote:
On 12/1/20 9:55 AM, Jeff Layton wrote:
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology:
https://tracker.ceph.com/issues/6373
It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts.
Is it possible to do this in teuthology? How would I approach this?
I think I can make this happen pretty easily in ceph-cm-ansible. What I'd need from you is desired filesystem type and size. Once I'm done with my end, we'll need to create a cephlab.ansible overrides yaml fragment to stick in that suite's qa directory.
Ok, cool: fstype: xfs or ext4 (either is fine) size: ~50g or so would be ideal, but we can probably get away with less if necessary Thanks! -- Jeff Layton <jlayton@redhat.com>
On 12/1/20 10:24 AM, Jeff Layton wrote:
On Tue, 2020-12-01 at 10:22 -0500, David Galloway wrote:
On 12/1/20 9:55 AM, Jeff Layton wrote:
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology:
https://tracker.ceph.com/issues/6373
It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts.
Is it possible to do this in teuthology? How would I approach this?
I think I can make this happen pretty easily in ceph-cm-ansible. What I'd need from you is desired filesystem type and size. Once I'm done with my end, we'll need to create a cephlab.ansible overrides yaml fragment to stick in that suite's qa directory.
Ok, cool:
fstype: xfs or ext4 (either is fine) size: ~50g or so would be ideal, but we can probably get away with less if necessary
Getting there... https://github.com/ceph/ceph-cm-ansible/pull/592 Can you give me an optimal cachefilesd.conf and I'll set that up in ceph-cm-ansible too?
On Tue, 2020-12-01 at 14:23 -0500, David Galloway wrote:
On 12/1/20 10:24 AM, Jeff Layton wrote:
On Tue, 2020-12-01 at 10:22 -0500, David Galloway wrote:
On 12/1/20 9:55 AM, Jeff Layton wrote:
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology:
https://tracker.ceph.com/issues/6373
It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts.
Is it possible to do this in teuthology? How would I approach this?
I think I can make this happen pretty easily in ceph-cm-ansible. What I'd need from you is desired filesystem type and size. Once I'm done with my end, we'll need to create a cephlab.ansible overrides yaml fragment to stick in that suite's qa directory.
Ok, cool:
fstype: xfs or ext4 (either is fine) size: ~50g or so would be ideal, but we can probably get away with less if necessary
Getting there...
https://github.com/ceph/ceph-cm-ansible/pull/592
Can you give me an optimal cachefilesd.conf and I'll set that up in ceph-cm-ansible too?
I never actually tweak the defaults in my local testing, so let's start with a default cachefilesd.conf as installed by the package. If we need to tweak it later, then we'll go from there. Thanks! -- Jeff Layton <jlayton@redhat.com>
On 12/1/20 3:07 PM, Jeff Layton wrote:
On Tue, 2020-12-01 at 14:23 -0500, David Galloway wrote:
On 12/1/20 10:24 AM, Jeff Layton wrote:
On Tue, 2020-12-01 at 10:22 -0500, David Galloway wrote:
On 12/1/20 9:55 AM, Jeff Layton wrote:
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology:
https://tracker.ceph.com/issues/6373
It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts.
Is it possible to do this in teuthology? How would I approach this?
I think I can make this happen pretty easily in ceph-cm-ansible. What I'd need from you is desired filesystem type and size. Once I'm done with my end, we'll need to create a cephlab.ansible overrides yaml fragment to stick in that suite's qa directory.
Ok, cool:
fstype: xfs or ext4 (either is fine) size: ~50g or so would be ideal, but we can probably get away with less if necessary
Getting there...
https://github.com/ceph/ceph-cm-ansible/pull/592
Can you give me an optimal cachefilesd.conf and I'll set that up in ceph-cm-ansible too?
I never actually tweak the defaults in my local testing, so let's start with a default cachefilesd.conf as installed by the package. If we need to tweak it later, then we'll go from there.
K. I made it optional and used the package-provided defaults. Your requested changes are implemented in ceph-cm-ansible. I've attached the yaml fragments you'll need to pass on the `teuthology-suite` command line to set up the partition, filesystem, and cachefilesd service. e.g., teuthology-suite ... ~/fs-cache-smithi.yml
On Tue, 2020-12-01 at 16:31 -0500, David Galloway wrote:
On 12/1/20 3:07 PM, Jeff Layton wrote:
On Tue, 2020-12-01 at 14:23 -0500, David Galloway wrote:
On 12/1/20 10:24 AM, Jeff Layton wrote:
On Tue, 2020-12-01 at 10:22 -0500, David Galloway wrote:
On 12/1/20 9:55 AM, Jeff Layton wrote:
I've been working on a patch series to overhaul the fscache code in the kclient. I also have this (really old) tracker to add fscache testing to teuthology:
https://tracker.ceph.com/issues/6373
It would be ideal if the clients in such testing had a dedicated filesystem mounted on /var/cache/fscache, so that if it fills up it doesn't take down the rootfs with it. We'll also need to have cachefilesd installed and running in the client hosts.
Is it possible to do this in teuthology? How would I approach this?
I think I can make this happen pretty easily in ceph-cm-ansible. What I'd need from you is desired filesystem type and size. Once I'm done with my end, we'll need to create a cephlab.ansible overrides yaml fragment to stick in that suite's qa directory.
Ok, cool:
fstype: xfs or ext4 (either is fine) size: ~50g or so would be ideal, but we can probably get away with less if necessary
Getting there...
https://github.com/ceph/ceph-cm-ansible/pull/592
Can you give me an optimal cachefilesd.conf and I'll set that up in ceph-cm-ansible too?
I never actually tweak the defaults in my local testing, so let's start with a default cachefilesd.conf as installed by the package. If we need to tweak it later, then we'll go from there.
K. I made it optional and used the package-provided defaults.
Your requested changes are implemented in ceph-cm-ansible. I've attached the yaml fragments you'll need to pass on the `teuthology-suite` command line to set up the partition, filesystem, and cachefilesd service.
e.g., teuthology-suite ... ~/fs-cache-smithi.yml
Many thanks! I now need to figure out how to pass arbitrary mount options to the kclient. I think that might need some plumbing down in ceph/qa. I'll sort that out soon and test this. Cheers! -- Jeff Layton <jlayton@redhat.com>
participants (2)
-
David Galloway
-
Jeff Layton