Upgrade the kafa nodes:
journal0.internal.staging.swh.network(staging's kafka was migrated to storage1, which can't be upgraded to bullseye as swh-storage and swh-objstorage are deployed on it)- kafka[1-4].internal.softwareheritage.org
Upgrade plan:
- disable puppet
- upgrade buster packages
- stop and disable kafka
- stop and disable zookeeper
- reboot if needed
- upgrade to bullseye following the plan described in T3749
- Try to reboot, if the network is not working
- edit the /etc/network/interfaces file and comment the physical interface declarations
auto enp94s0f0np0 iface enp94s0f0np0 inet manual mtu 9000 # interface enp94s0f1np1 configuration auto enp94s0f1np1 iface enp94s0f1np1 inet manual mtu 9000
- Remove the interface configuration on puppet (kafka[1-4].internal.softwareheritage.org` files) before restarting puppet
- restart and check zookeeper status
- restart and check kafka status (wait for the cluster to recover[1])
- re-enable puppet
[1] http://getty.internal.softwareheritage.org:9000/clusters/rocquencourt/topics