I'm trying to upgrade our 3-monitor cluster from Centos 7 and Nautilus to Rocky 9 and Quincy. This has been a very slow process of upgrading one thing, running the cluster for a while, then upgrading the next thing. I first upgraded to the last Centos 7 and upgraded to Octopus. That worked fine. Then I was going to upgrade the OS to Rocky 9 while staying on Octopus, but then found out that Octopus is not available for Rocky 9. So I broke my own rule and upgraded one of the monitor (and manager) nodes to Rocky 9 and Pacific, then rejoined it to the cluster. That seemed to work just fine. Feeling bold, I upgraded the second monitor and manager node to Rocky 9 and Pacific. That also seemed to work fine, with the cluster showing all the monitors and managers running. But now, if I shut down the last "Octopus" monitor, the cluster becomes unresponsive. This only happens when I shut down the Octopus monitor. If I shut down one of the Pacific monitors, the cluster keeps responding with the expected: "HEALTH_WARN 1/3 mons down" and then goes back to normal when the monitor process is started again. Is this expected? What am I missing? Thanks for any pointers! _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |