Cleanup #36736
Lvm disks are not cleaned up during recreation of nodes for new cluster
0%
Description
If there are disks with existing lvm partition, this is not cleaned up before creating a new cluster.
As a result stale data might be present on the disk and this is interfering with new cluster.
In my case I am using ceph-volume lvm to create partitions in every run, every new run which uses a node with previously used lvm disks faces some problems because of stale entries in earlier lvm partitions, so this needs to be cleaned up at the beginning of node creation.
History
#1 Updated by Vasu Kulkarni over 5 years ago
- Assignee set to David Galloway
- Priority changed from Normal to High
Since we are setting up cluster with lvm's as well, I think its better to handle cleanup in cm-ansible.
#2 Updated by David Galloway over 5 years ago
Are you creating these clusters manually or with teuthology?
ceph-cm-ansible is supposed to already clean up old LVMs. https://github.com/ceph/ceph-cm-ansible/blob/master/roles/testnode/tasks/zap_disks.yml#L68
#3 Updated by Shylesh Kumar over 5 years ago
Hi David,
We are using a modified ceph task in teuthology, which uses ceph-volume lvm to provision the osds.
For the first time(when disk has non-lvm) it works fine and any subsequent osd creation on same disk will see some stale data.
Thanks,
Shylesh
David Galloway wrote:
Are you creating these clusters manually or with teuthology?
ceph-cm-ansible is supposed to already clean up old LVMs. https://github.com/ceph/ceph-cm-ansible/blob/master/roles/testnode/tasks/zap_disks.yml#L68
#4 Updated by David Galloway over 5 years ago
Can you show me a job or pastebin example?
#5 Updated by Shylesh Kumar over 5 years ago
It was bug in my teuthology task itself, my bad. closing this as not a bug.
David Galloway wrote:
Can you show me a job or pastebin example?
#6 Updated by David Galloway about 5 years ago
- Status changed from New to Closed