Not sure if this is the appropriate place to report this and if this is specific to our site, but this merge breaks infiniband configuration on our Rocky 9 cluster.
The error response we got in the script logs for configib is:
"nmcli Error: unknown connection"
Digging into this, the IB looks to never have been connected via "nmcli con add " before calls to "nmcli con modify" are made.
I modified the configib script after Line 491 to make the following call instead of trying to write an ifcfg-$nic file.
elif [ $OS_name == 'redhat' ]
then
# First ip address
if [ $ipindex -eq 1 ]
###
if [[ "$OSVER" =~ (rhels9|alma9|rocky9) ]]; then
prefix=$(convert_netmask_to_cidr $netmask)
nmcli con add type infiniband con-name $nic ifname $nic ipv4.method manual ipv4.addresses $nicip/$prefix connection.autoconnect-priority 9
else
...
fi
....
I can possibly work on creating a pull request to address this, but it might take me awhile as I am a bit short on free time.
As far as how to approach the problem, it is not clear to me if copying the configipv4 and configipv6 function in the configeth script with the appropriate changes would be a good approach to take as I would assume those functions have had extensive work already done to cover many of the cases one would encounter with xcat.
Originally posted by @landcaster in #7444 (comment)
Originally posted by @landcaster in #7444 (comment)