TrueNAS Scale Linux mode 1 Bond

[hp_inkjet]

Hi,
Yesterday trying out TrueNAS Scale, on test DL380G10 with Connect-X 5 and LSI 3008 HBA, I found out that my 80G Active-Backup (linux mode 1) bond didn't work as expected.

I pulled out one cable from the NIC for maintenance and sensu started complaining about the host reachability, after some test I found out that my mode 1 bond didn't have a miimon rate set (cat /proc/net/bonding/bond0). Thus the bond couldn't understand when the carrier is down and is time to switch slave

I started digging and I found the python file that create the bond reading the configuration from the DB, the net frontend used is iproute2 and the command line seems right, minus the missing miimon part, I've patched the file as attached and now my bond works as expected.

I hope that this issue can be solved easily and this post can help someone else with my same problem

TrueNAS scale is a great product and you guys rocks,
Matteo

 

[Kris Moore]

@hp_inkjet - Glad to hear you are enjoying TrueNAS SCALE, and good job on figuring out how to patch that!

To prevent your work from getting lost on the forums, would you mind opening a ticket with the patch on jira.ixsystems.com, or perhaps even a pull request into https://github.com/truenas/middleware ? That'll ensure that this fix gets included in future updates of TrueNAS SCALE for you and other users who have similar setups.

 

[hp_inkjet]

Done!
Thank you Kris

 

[ZFSLover]

Is there a fix for this to set miimon for active/backup bonds? miimon is not set on these sorts of bonds, so when a link goes offline, it never fails over.

 

[oh2ftu]

I've encountered the same. The failover does not work as one would imagine.
Both interfaces (Mellanox ConnectX-3 and integrated intel i219) show

Code:

MII Status: up

Even if either is disconnected. Hence, it does not failover.
Speed does show Unknown while disconnected.
I did create a new bond and moved all interfaces over - it did not help.
I'm running Scale 22.03.3