Configure Ceph to mark flapping OSDs out of the cluster. Flapping OSD impact performance.
Issue
Configure Ceph to permanently mark flapping OSDs out of the cluster.
A marginal
failing
flapping
disk can impact system performance
How to prevent a marginal
or failing
or bad
OSD from restarting endlessly (always), which can effect system performance
(latency).
The systemd file defining how the OSD service is handled has Restart = always
, with no other limits defined, which may result in an endless restart loop of a marginal disk
For reference, see Section 7.1 of the Ceph 4.2 Release Notes, see the link for BZ #1860739 below.
Environment
Red Hat Ceph Storage (RHCS) 4.x
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.