`kubectl rook-ceph rook purge-osd` command fails

[tasgon]

A node in our cluster went down and had to be wiped, and I'm now trying to remove the OSDs that were on it. Unfortunately, I keep on running into this error:

% kubectl rook-ceph rook purge-osd 109
Info: Running purge osd command
2025/04/23 00:59:31 maxprocs: Updating GOMAXPROCS=16: determined from CPU quota
2025-04-23 00:59:31.433189 W | cephcmd: loaded admin secret from env var ROOK_CEPH_SECRET instead of from file
2025-04-23 00:59:31.433239 I | rookcmd: starting Rook v1.16.4 with arguments 'rook ceph osd remove --osd-ids=109 --force-osd-removal=false'
2025-04-23 00:59:31.433245 I | rookcmd: flag values: --force-osd-removal=false, --help=false, --log-level=INFO, --osd-ids=109, --preserve-pvc=false
2025-04-23 00:59:31.433249 I | ceph-spec: parsing mon endpoints: e=<ip1>:6789,j=<ip2>:6789,h=<ip3>:6789,a=<ip4>:6789,b=<ip5>:6789
2025-04-23 00:59:31.443860 I | cephclient: writing config file /var/lib/rook/rook-ceph/rook-ceph.config
2025-04-23 00:59:31.444004 I | cephclient: generated admin config in /var/lib/rook/rook-ceph
2025-04-23 00:59:31.932175 C | rookcmd: failed to get osd dump: failed to unmarshal osd dump response: invalid character 'i' looking for beginning of value
Error: . failed to run command. command terminated with exit code 1%!(EXTRA string=failed to remove osd %s, string=109)

(I've omitted the mon IPs.)

This is happening with the v0.9.3 of the Krew plugin and I've tried with and without --force. Could this be fixed by upgrading our Rook version? I looked through the changelog and didn't see anything obvious about this.

[sp98]

@tasgon can you post the out ceph osd dump --format json here?

[sp98]

@tasgon checking about my last comment.