[SOLVED] Setting up ceph, getting an error in the gui

[UdoB]

Im not sure but I guess you cant use the same network 192.168.0.0/24 for both ceph networks.

Well, yes you can - although there are really good reasons for two physically separated networks. I am testing it currently with

~# grep _network /etc/ceph/ceph.conf
         cluster_network = 10.3.16.7/16
         public_network = 10.3.16.7/16

...and it seems to work. Again: not recommended. And additionally: keep an eye on corosync, I have that one on different network cable.

Best regards

 

[jsterr]

Well, yes you can - although there are really good reasons for two physically separated networks. I am testing it currently with

~# grep _network /etc/ceph/ceph.conf
         cluster_network = 10.3.16.7/16
         public_network = 10.3.16.7/16

...and it seems to work. Again: not recommended. And additionally: keep an eye on corosync, I have that one on different network cable.

Best regards

Im curious, why are you testing this?

 

[rsr911]

Nice! Its ok I always like connecting to people, so if you want you can add me on linkedin (business) if you have! Greetings

Request sent. I just use the free account.

Is changing the ceph networks more involved than just changing the IPs now that it's up and working? I suppose it doesn't matter one of mine is a public IP (oops) as it's on an isolated switch.

 

[UdoB]

Im curious, why are you testing this?

Because I can ;-)

I am definitely not a Ceph specialist, I use ZFS primarily. This is just a homelab: some HM80 have a SATA/ZFS-Mirror for VM-data. Then there is a single build-in NVMe which I can not find a good way for to utilize. So I set up Ceph with a single OSD per node. (Three nodes to start with, probably five nodes for Christmas...)

And to be clear once again: this setup is definitely not recommended as it is probably just too small for a stable system - I am just gathering experience.

Best regards

 

[jsterr]

Because I can ;-)

I am definitely not a Ceph specialist, I use ZFS primarily. This is just a homelab: some HM80 have a SATA/ZFS-Mirror for VM-data. Then there is a single build-in NVMe which I can not find a good way for to utilize. So I set up Ceph with a single OSD per node. (Three nodes to start with, probably five nodes for Christmas...)

And to be clear once again: this setup is definitely not recommended as it is probably just too small for a stable system - I am just gathering experience.

Best regards

Nice! So you did not get the error from this post like @rsr911 got?

 

[UdoB]

Nice! So you did not get the error from this post like @rsr911 got?

Correct. For me it works with public_/cluster_network being the same address on the same network.

 

[rsr911]

I'd like to thank you guys for the help. My cluster is up and running and Ive got PBS on a separate PVE server doing backups. Hoping to order a fifth server this coming week so I can have a 5 node cluster (4 as data).

I'm going to call this issue solved. I'm about to post another, likely easier question as a new thread.