Forums

Home / Forums

You need to log in to create posts and topics. Login · Register

Setup Management NIC bonding after deployment


JediMasterSeamus
3 Posts
May 1, 2024, 9:45 pm

Hello there!

I've built a 5 node cluster - 3 management VMs (on 3 different hyper-v hosts), 2 physical boxes with storage.

I'd like to have redundant networking among our 2 10g switches.
Management gets 2 ports, Sync gets 2 ports, iSCSI1 gets 1 and iSCSI2 gets 1.
I am looking to do an active/backup bond for Management and Sync on the physical nodes.
Adding extra network adapters to the virtual nodes is not an issue if that's what's needed to get the bond.
I'm just not sure how to do it, since I've already deployed the cluster with the below networking.
(I can rebuild it if necessary, but it just means waiting for everything to reimage and I'd like to avoid that if possible).

Ethernet configuration on the physical nodes is -
eth0 - management
ethX0 - needs to be bonded to eth0, currently not doing anything
eth1 - sync/heartbeat
ethX1 - needs to be bonded to eth1, currently not doing anything
eth2 - iSCSI1
eth3 - iSCSI2

Ethernet configuration on the virtual nodes is -
eth0 - management
eth1 - sync/heartbeat
eth2 - iSCSI1
eth3 - iSCSI2

Thank you in advance for any help!

-JMS

#1

admin
2,930 Posts
May 1, 2024, 10:39 pm

You can create bond for management network using ui. it is best to have it in active/backup mode to avoid disconnections while being configured. typically you would use lacp for backend or i/o networks.

You can manually edit the configuration after deployment in file
/opt/petasan/config/cluster_info.json
the syntax is self explanatory, but you can also create a dummy test and deploy 1 vm with the configuration/bonds you wish and use the configuration in an already built cluster.

You can perform any customisation, re-configuration in a custom script
/opt/petasan/scripts/custom/post_start_network.sh
but generally editing the json file will be more than enough.

#2

JediMasterSeamus
3 Posts
May 1, 2024, 11:47 pm

Alright, I was able to get the info I needed from the test cluster.
Thank you for the direction on that, it's what I needed to see.

Now, if I make a change to the cluster_info.json, will that reflect across all nodes, or just the one I made the change to?

#3

JediMasterSeamus
3 Posts
May 2, 2024, 4:47 am

Actually, I figured that bit out, and it's all working now.
In case anyone finds this later:
Pull the file from your dummy cluster and your live cluster with scp.
Edit to fit, then scp it back to your live cluster.
Then reboot the whole thing for the config to take.

#4