that means that creating a temporary file in /var/lib/ceph/mgr/ceph-bowie failed. did the directory creation actually succeed? (i.e., does /var/lib/ceph/mgr/ceph-bowie exist?)
New Ceph 12.1.1 packages for testing, GUI for creating bluestore OSDs
- Thread starter martin
- Start date
that means that creating a temporary file in /var/lib/ceph/mgr/ceph-bowie failed. did the directory creation actually succeed? (i.e., does /var/lib/ceph/mgr/ceph-bowie exist?)
Neither that nor /var/lib/ceph/mgr does exist.
It seems that by some magical reason the ceph-mgr package doesn't work unless it's actually installed. :-]
The moral of the story, though, is that pveceph should verify the directory existence (and thus the existence of the installed package) before acting.
could you file a bug for this? (feel free to generalize it a bit, as it is not only ceph-mgr that is hit by this). thanks!
Hi,
Seems that 12.1.4 is available and fix some upgrade issues on 12.1.3....wait and see
What will happen to the ones like me who have installed Luminous RC on pve 5 (when stable version is ready )? We must switch to the official repository or keep yours ?
Thanks again for your work.
Antoine
Seems that 12.1.4 is available and fix some upgrade issues on 12.1.3....wait and see
What will happen to the ones like me who have installed Luminous RC on pve 5 (when stable version is ready )? We must switch to the official repository or keep yours ?
Thanks again for your work.
Antoine
Hi,
Seems that 12.1.4 is available and fix some upgrade issues on 12.1.3....wait and see
What will happen to the ones like me who have installed Luminous RC on pve 5 (when stable version is ready )? We must switch to the official repository or keep yours ?
Thanks again for your work.
Antoine
we will provide stable packages for Proxmox VE via download.proxmox.com
Thanks.
I must admit that performance with Luminous and Bluestore is really amazing, even with only few sas drives...
I must admit that performance with Luminous and Bluestore is really amazing, even with only few sas drives...
I've been testing PVE 5.0 with ceph 12.1.x (luminous release candidate).
•cephfs on ec pool seems OK; want to do a bit more testing, but performance is pretty good
•rbd on ec pool doesn't work yet from the GUI — the CLI requires a new option --data-pool, which specifies which ec pool to actually store the data in; a replicated pool provides the metadata, and that's the pool that would be listed in storage.cfg. So it looks like we need some reworking of the underlying rbd commands in order to be able to use the GUI.
I'll test manually migrating a VM (creating the files on a different pool first).
•cephfs on ec pool seems OK; want to do a bit more testing, but performance is pretty good
•rbd on ec pool doesn't work yet from the GUI — the CLI requires a new option --data-pool, which specifies which ec pool to actually store the data in; a replicated pool provides the metadata, and that's the pool that would be listed in storage.cfg. So it looks like we need some reworking of the underlying rbd commands in order to be able to use the GUI.
I'll test manually migrating a VM (creating the files on a different pool first).
Hi,
Seems that 12.1.4 is available and fix some upgrade issues on 12.1.3....wait and see
What will happen to the ones like me who have installed Luminous RC on pve 5 (when stable version is ready )? We must switch to the official repository or keep yours ?
Thanks again for your work.
Antoine
12.1.4 is available in the 'test' repository at http://download.proxmox.com/debian/ceph-luminous/dists/stretch/test/binary-amd64/
remove the storage entry, stop/remove/destroy all OSDs, managers, and monitors (in that order), remove any leftover config files and directories in /etc/ceph/ and /var/lib/ceph/**/ (don't remove the /etc/ceph and /var/lib/ceph/* directories themselves though!). remove the ceph-related files in /etc/pve/priv/ . if you want, you can then remove all the ceph-* packages except ceph-common, but that is not really needed (without a config and without any service instances, the packages don't do anything).
DISCLAIMER: all of the above is from memory and not tested please try it out on a test system before attempting anything like this on a production system.
DISCLAIMER: all of the above is from memory and not tested
please try it out on a test system before attempting anything like this on a production system.we have a production ceph storage in use. I want to set up another ceph cluster to test before upgrading.
we use an unmanaged 10G switch for ceph .
Is it OK to use the same network IP subnet for ceph on production and test? or might that mess things up?
we use an unmanaged 10G switch for ceph .
Is it OK to use the same network IP subnet for ceph on production and test? or might that mess things up?
- pveceph install --version luminous makes a ceph.list with wrong sources. so had to do this to get pvetest debs:
then after install
Code:
echo 'deb http://download.proxmox.com/debian/ceph-luminous stretch test' > /etc/apt/sources.list.d/ceph-test.list
Code:
rm /etc/apt/sources.list.d/ceph.list
Last edited:
Is it correct that setting tunables to optimal doesn't work?
root@nod1:~/ceph# ceph osd crush tunables optimal
Error EINVAL: new crush map requires client version jewel but require_min_compat_client is firefly
root@nod1:~/ceph# ceph osd crush tunables optimal
Error EINVAL: new crush map requires client version jewel but require_min_compat_client is firefly
It's a protection if you have still older client connected. (old librbd for qemu, old kernel for lxc, maybe some stuck "rbd ..." command launched previously by proxmox,....)
Is there any plan to promote release the 12.1.4 RC to the regular updates instead 12.1.2?
I want to mention that I really appreciate Proxmox, because their staff and the community.
Thanks
to the regular updates instead 12.1.2? I want to mention that I really appreciate Proxmox, because their staff and the community.
Thanks