pricing and other dummy questions

dinkohrzica

New Member
May 9, 2014
10
0
1
I have 2 cpu server where I have bought 2 cpu licence. To have 2 node HA I need another 2cpu licence I guess?
I do like 2 node HA instead of 3 node HA so can I ?
1. run software raid on 2 node HA?
2. can I deploy 2 more dedicated servers with two node HA and then have proper quorum? or? I am bit confused with 4 dedicated servers both on 2 node HA :)

Thanks a lot for clarification.
Regards!
 
Hi,
I have 2 cpu server where I have bought 2 cpu licence. To have 2 node HA I need another 2cpu licence I guess?
the license has nothing to do with 2-node-ha - you need for the enterprise repository an subscrition for each cpu socket in the cluster.
If you node 1 has two sockets and your node 2 only one you will need an 2-socket subscrition plus an one-socket-subscrition.
I do like 2 node HA instead of 3 node HA so can I ?
1. run software raid on 2 node HA?
software raid has also nothing to do with 2-node-ha and is not supported by the devs (but some people here use softwareraids).
2. can I deploy 2 more dedicated servers with two node HA and then have proper quorum? or? I am bit confused with 4 dedicated servers both on 2 node HA :)
Thanks a lot for clarification.
Regards!
I don't recommend only two nodes with an cluster - I guess you will run intro trouble (and this first, if you don't need additional trouble, because you have enough if one node die).

Udo
 
Thanks for all clarification. So 3 nodes is a must. So now I have 2 servers directly connected with 2nd network interface (without a switch). Can I add one more server and connect it in this way:
1 server is connected to internet and directly to server 2
2nd server is directly connected to server 1 and with server 3
3rd server is connected to internet and to server 2 directly

If node 1 fails (where all will be run including ceph) I wish to start all servers on node 3 (can I put it like this?)
Why do I wish to connect servers like that it it becouse I do not wish to use a switch at this time. As those are 1 gbit network cards is it enough for ceph?

Thanksa lot guys :)
 
theorically, with this setup meant server1 and server3 only can communicate with each other using internet. How fast is your connection speed then? I doubt it to be fast enough since the recommended connection speed is 10gbit network. You should get a gigabit switch and make use of the 2 nics in the servers which could yield 2gbit combined transfer. At least better speed than 1gb and certainly faster than relying on internet connection. Then again, I am a noob myself and commenting based on what I read about ceph so far.
 
Thanks for all clarification. So 3 nodes is a must. So now I have 2 servers directly connected with 2nd network interface (without a switch). Can I add one more server and connect it in this way:
1 server is connected to internet and directly to server 2
2nd server is directly connected to server 1 and with server 3
3rd server is connected to internet and to server 2 directly

If node 1 fails (where all will be run including ceph) I wish to start all servers on node 3 (can I put it like this?)
Why do I wish to connect servers like that it it becouse I do not wish to use a switch at this time. As those are 1 gbit network cards is it enough for ceph?

Thanksa lot guys :)
Hi,
this setup will not work. In an cluster all nodes must be in one network (multicast) - and all must reach all others.

And the same is with ceph-nodes (here is an second seperate network not bad)

Udo
 
Without switch you are only able to connect two servers.
thanks a lot for everyone who answered...
well with just 2 node setup is it better to have HA or without HA? Can things go bad if I have just 2 regular nodes?
Like 2 nodes with 2 disks in software radi (I read somewhere that its not recommended not sure why)
 
thanks a lot for everyone who answered...
well with just 2 node setup is it better to have HA or without HA? Can things go bad if I have just 2 regular nodes?
Like 2 nodes with 2 disks in software radi (I read somewhere that its not recommended not sure why)
Hi,
you can use 2 nodes with drbd-storage (also 2 drbd-ressources - one for each node).

If one node fail, you automaticly lost quorum.
But you can manual reach quorum again (set expected=1) and then move the VM-config from the failed node to the remaining node (like "mv /etc/pve/nodes/FAILEDNODE/qemu-server/* /etc/pve/nodes/RUNNINGNODE/qemu-server/") and start the VMs there.

Udo
 
Hi,
you can use 2 nodes with drbd-storage (also 2 drbd-ressources - one for each node).

If one node fail, you automaticly lost quorum.
But you can manual reach quorum again (set expected=1) and then move the VM-config from the failed node to the remaining node (like "mv /etc/pve/nodes/FAILEDNODE/qemu-server/* /etc/pve/nodes/RUNNINGNODE/qemu-server/") and start the VMs there.

Udo

Thanks Udo, the thing is everybody cry's do not use 2 node HA...so I am confused is it really a mess or I can easy manually just say=use 2nd node as alive and thats it???
I am not sure if this:
you can use 2 nodes with drbd-storage (also 2 drbd-ressources - one for each node).
means that I can use HA and not HA vps's inside my 2 nodes/dedicated server?
Again thanks :)