cephadm: check "ceph_device_lvm" field instead of "ceph_device" during zap

This field was renamed as part of
https://github.com/ceph/ceph/commit/4941d098e337f2b7ad8c6f7c90be3ae252d22f7b
but the cephadm piece was still looking for "ceph_device" meaning
we would never zap any devices. Worth noting this won't
get caught by teuthology as we make OSDs in that env directly
on top of LVs made out of a split up nvme drive and
ceph-volume inventory doesn't list LVs.

Signed-off-by: Adam King <adking@redhat.com>
(cherry picked from commit fd8a05b2cb97370fe770ca18b50ad1c3ab763447)

diff --git a/src/cephadm/cephadm.py b/src/cephadm/cephadm.py
index 6854fa22b231063dd3bd3d37b0f2b1f75224599e..5758e7c258b363c9719961c6073ef3fcce09218e 100755 (executable)
--- a/src/cephadm/cephadm.py
+++ b/src/cephadm/cephadm.py
@@ -4290,7 +4290,7 @@ def _zap_osds(ctx: CephadmContext) -> None:
         raise Error(f'Invalid JSON in ceph-volume inventory: {e}')
 
     for i in ls:
-        matches = [lv.get('cluster_fsid') == ctx.fsid and i.get('ceph_device') for lv in i.get('lvs', [])]
+        matches = [lv.get('cluster_fsid') == ctx.fsid and i.get('ceph_device_lvm') for lv in i.get('lvs', [])]
         if any(matches) and all(matches):
             _zap(ctx, i.get('path'))
         elif any(matches):