Hello inverse team ! I'm having a few issues upgrading my 3 servers cluster from version 12.2.0 to 13.1.0. I'll explain the process I followed and what issues I've encountered.
I first read the "Clustering Quick Installation Guide", part 12.4 "Performing an upgrade on a cluster" as mentionned in the Upgrade Guide, part 5.2. I do my backup via snapshots and start the process. Disabling the auto correction and galera autofix goes well. I can detach the Node C and make A and B run smoothly. The company network and service aren't interrupted. I then proceed to upgrade node C with the following command : /usr/local/pf/addons/upgrade/do-upgrade.sh During the process, the script showed me an error : Moving conf/switches.conf.dpkg-dist -> conf/switches.conf Attempting a dry-run on the patch on conf/switches.conf checking file conf/switches.conf Hunk #1 FAILED at 1. 1 out of 1 hunk FAILED Patching conf/switches.conf failed. I re run the script and it goes "smoothly". I run the checkup and no errors and proceed to switch production between Node A, B and C. With node C taking the HA IP, we still couldn't authenticate on the network. I check the config and all the switches were gone ! I revert from my snapshot and proceed the same until the switch error. I open the file and check the content. It's empty. It seems like it got replaced by the default config file so I tried to copy back the backup with all my previous switches configuration and then run the checkup : systemctl start packetfence-proxysql /usr/local/pf/bin/pfcmd checkup I get a Warning saying the switch type "Cisco::Catalyst_2960" don't exist. So I replaced all instance of the file with the type "Cisco::IOS_15.0" I run the checkup again and it output no warning or errors. To avoid service interruption, I've kept the node A and B up with the C too. A and B where together and C still in standalone. I could authenticate with a switch on node A, B and the HA IP but I couldn't authenticate on node C. I've tried many PF configuration to make it work but it just won't. Packetfence.log output nothing. Radius.log output nothing. It have the same behaviours as the switch wasn't declared in PF (but it was). To this point I don't know how can I go further into the update without going with a new cluster from scratch. Thanks a lot for your help. Greats, Adrian DESSAIGNE Enregistrer Enregistrer
_______________________________________________ PacketFence-users mailing list PacketFence-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/packetfence-users
