18 May
2020
18 May
'20
12:55 a.m.
Clarification: On Sun, May 17, 2020 at 10:36:41PM +0200, Martin Millnert wrote:
Hi Jeff,
Ran into the very same issue. Filed a bug-report at https://tracker.ceph.com/issues/45574 ceph 15.2.1 on up-to-date Debian Buster.
TL;DR: The way ceph-mgr-rook's RookOrchestrator class interacts with python3-numpy package is borked. Result: Cluster cannot start, since 'devicehealth' plugin of ceph-mgr is an always-on module.
Mistaken cause for cluster not starting -- cluster came from Luminous through Nautilus to Octopus and had a ceph osd require-osd-release = luminous, preventing OSD's from launching. Now cluster happy Octopus, but Python bug remain. /M