Hi Marc, While it might be tempting to rely on Ceph itself to store the monitoring data, IMHO it would be a bad idea: the monitoring I/O load might interfere with the cluster performance, and, in a worst case scenario, the monitoring data might become unavailable (probably the exact kind of situation where you most needed those data). The current monitoring stack HA for stateful services is based on statically-pinned redundant instances (e.g.: ceph orch apply prometheus --placement="monitoring[1-3]"). For stateless ones there's work planned to extend the Cephadm scheduler to support dynamic rescheduling ( https://tracker.ceph.com/issues/47038). Kind Regards, Ernesto On Fri, Nov 26, 2021 at 10:42 AM Marc <Marc@xxxxxxxxxxxxxxxxx> wrote: > > - Now, with Cephadm, you can easily deploy a highly-available > > monitoring > > Really? Sorry for asking because I don't know anything about podman. But > does podman automatically start tasks on a different node when one is down? > And do they have some native ceph support for external volumes to > facilitate statefull tasks or are they using this 'buggy' csi-ceph driver? > > > > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |