Open f0 opened 6 years ago
We found the reason (maybe?) We do not use bond0, we disable this with
cat /etc/modprobe.d/bonding.conf
options bonding max_bonds=0
and then systemd creates the netdev bonding.
With this setup, the problem happen
When we change to bond0 (remove /etc/modprobe.d/bonding.conf ) and rename the netdev to bond0, it works, the problem is gone (But Mode is RR not LACP wich is what we want)
Using options bonding max_bonds=0 and manually configuring bonding (ip link) also works
To generalize:
Mode Round Robin does Work, Mode 802ad (LACP) is broken
it turns out that max_bonds=0 is the default setting that comes with systemd and after removing up/down delay, and mii-monitor from the bonding.netdev, more systems are stable , but not all....
LACPTransmitRate=fast solves this Problem for us......
Any detail? We are having similar issues
Issue Report
Bug
Container Linux Version
Environment
Dell Inc. PowerEdge R630/0CNCJW,
Expected Behavior
System has stable network connectivity
Actual Behavior
After the update (1409.7.0 to 1520.6.0) ,there is no stable network connectivity, systems does not even boot into the update partition, they revert to the previous state
With updelay set to 0 (tip fromhttps://github.com/coreos/bugs/issues/2065 ) they boot into the updated partition, then they are up for a short time, and then the network breaks and the error seen in https://github.com/coreos/bugs/issues/2065 happens (seen on the HW console)
Reproduction Steps
Update from 1409.7.0 to 1520.6.0 (Linux 4.11.11-coreos to 4.13.5-coreos)
Other Information
is similar to https://github.com/coreos/bugs/issues/2065
lspci