After (successfully!) converting a single machine ceph to a multi machine ceph, I can confirm this is what any interaction with your ceph cluster and the ceph documentation feels like
Though I have not changed the CRUSH map (the algorithm that assigns objects/replicas to disks) from "different disk" to "different physical machine" yet, that will be a chaotic moment
After countless times creating a single-host ceph cluster, then adding a second monitor and watching it always go in an endless "probing" state… I stopped trying. 😭 Probably Cassandra set my bar too high, but why is any other cluster DB or filesystem so complex and fickle? (I tried with 3 different versions, all nodes reach each other on both needed ports, zero errors in the logs even at 10/10 level… I did it all, again and again, including RTFS)
benjojo
in reply to benjojo • • •You love to see it!
Though I have not changed the CRUSH map (the algorithm that assigns objects/replicas to disks) from "different disk" to "different physical machine" yet, that will be a chaotic moment
Lapo Luchini
in reply to benjojo • •After countless times creating a single-host ceph cluster, then adding a second monitor and watching it always go in an endless "probing" state… I stopped trying. 😭
Probably Cassandra set my bar too high, but why is any other cluster DB or filesystem so complex and fickle?
(I tried with 3 different versions, all nodes reach each other on both needed ports, zero errors in the logs even at 10/10 level… I did it all, again and again, including RTFS)
I hope you cluster will work better. 💪🏼
Igel likes this.
benjojo
in reply to Lapo Luchini • • •Lapo Luchini likes this.