osd: heartbeat can't keep up with large cluster changes
in wido's case, a new crushmap makes osds flap.
#2 Updated by Sage Weil over 9 years ago
Greg Farnum wrote:
Do we still think this is an issue after 856999eda434fa9b7d93b152427cf7c82240f220 ("osd: clear failure_queue when marked down"), or were there other issues with OSD crushmap changes that started the chain and the delayed failure_queue just kept it going?
There might be multiple issues, not sure. I want to make sure it's working well under pretty heavy osd repeering load before closing this out.