XOSTOR hyperconvergence preview
-
@Theoi-Meteoroi said in XOSTOR hyperconvergence preview:
You lost quorum.
Not a quorum issue:
exists device name:xcp-persistent-database volume:0 minor:1000 backing_dev:/dev/linstor_group/xcp-persistent-database_00000 disk:UpToDate client:no quorum:yes
-
@Maelstrom96 Thank you for the logs, I'm trying to understand the issue.
For the moment I don't see a problem regarding the status of the services. -
@Maelstrom96 It sounds like a race condition or a bad mount of the database. But I'm not sure, so I will add more logs for the next RPM. We plan to release it in a few weeks.
-
@ronan-a I will be testing my theory a little bit later today, but I believe it might be a hostname mismatch between the node name it expects in linstor and what it set to now on Dom0. We had the hostname of the node updated before the cluster was spinned up, but I think it still had the previous name active when the linstor SR was created.
This means that the node name doesn't match here:
https://github.com/xcp-ng/sm/blob/e951676098c80e6da6de4d4653f496b15f5a8cb9/drivers/linstorvolumemanager.py#L2641C21-L2641C41I will try to revert the hostname and see if it fixes everything.
Edit: Just tested and reverted the hostname to the default one, which matches what's in linstor, and it works again. So seems like changing a hostname after the cluster is provisionned is a no-no.
-
@Maelstrom96 Oh! This explanation makes sense, thank you. Yes in case of change of hostname, the LINSTOR node name must also be modified, otherwise the path to the database resource will not be found.
-
@ronan-a Do you know of a way to update a node name in Linstor? I've tried to look in their documentation and checked through CLI commands but couldn't find a way.
-
@Maelstrom96 Well there is no simple helper to do that using the CLI.
So you can create a new node using:
linstor node create --node-type Combined <NAME> <IP>
Then you must evacuate the old node to preserve the replication count:
linstor node evacuate <OLD_NAME>
Next, you can change your hostname an restart the services on each host:
systemctl stop linstor-controller systemctl restart linstor-satellites
Finally you can delete the node:
linstor node delete <OLD_NAME>
After that you must recreate the diskless resources if necessary. Exec
linstor advise r
to see the commands to execute. -
@ronan-a Thanks a lot for that procedure.
Ended up needing to do a little bit more, since for some reason, "evacuate" failed. I deleted the node and then went and just manually recreated my resources using:
linstor resource create --auto-place +1 <resource_name>
Which didn't work at first because the new node didn't have a storage-pool configured, which required this command to work (NOTE - This is only valid if your SR was setup as thin):
linstor storage-pool create lvmthin <node_name> xcp-sr-linstor_group_thin_device linstor_group/thin_device
Also, worth nothing that before actually re-creating the resources, you might want to manually clean up the lingering Logical Volumes that weren't cleaned up if evacuate failed.
Find volumes with:
lvdisplay
and then delete them with:
lvremove <LV Path>
example:
lvremove /dev/linstor_group/xcp-persistent-database_00000
-
I've using xcp-ng with NFS Shared Storage for some months now and I am happy with it so far.
I've some ssds in a 3 server setup and I'd like to test xostor. Before I will setup xostor, there are some questions regarding it, as I am only familiar with virtuozzo/acronis-storage and ceph so far. Are there the same restrictions for using xostor, that exists in ceph e.g.?- only enterprise ssds because of power loss protection
- do not use Raids ( especially Raid 0 ) if the controller is capable of using HBA Mode. I've an Dell H330 Controller and if Raid is no problem, I'd like to setup OS with Hardware Raid 1 and xostor on the rest of the ssds with raid0 arrays per each disk. If HBA mode is prefered, I need to stick with Software Raid 1, I think. Software Raid 1 is working fine, but I've had some problems in the past if the boot drive of the mirror died...
- I've installed xen-orchestra manually. Once xostor is installed, will the xostor button in xen orchestra will have a function or is it only available within the XOA appliance?
Thanks for your answers!
-
- That shouldn't be a problem, if it's bad on one disk for one host, it should be resync. It's not a big filesystem shared, it's only a block space split between each created virtual disk.
- That should work fine (@ronan-a will confirm but I don't think there's very low level optimization that could be affected by a RAID card?)
- XOSTOR UI is only available in XOA, but you'll be able to manage and have all the features from the CLI
-
@flibbi, @olivierlambert RAID shouldn't be problem for XOSTOR. During some of my testing shortly after the preview was released, I was running it on software RAID 10 arrays on each of my test servers. As long as the RAID isn't some sort of "fake RAID" and is done in hardware, it should work fine.