Galera garbd terminated: Group requested gcs_proto_ver: 2, max supported by this node: 0

Written by Claudio Kuenzler - 0 comments

Published on May 31st 2023 - Listed in Galera Database MariaDB

---

After upgrading data nodes of a Galera cluster from MariaDB 10.3 to 10.4 and from Galera-3 to Galera-4, an Arbitrator server was unable to sync with the cluster.

The garbd log file showed the following error:

2023-05-31 15:21:44.101  INFO: Quorum results:
    version    = 6,
    component  = PRIMARY,
    conf_id    = 6,
    members    = 2/3 (joined/total),
    act_id     = 149331040,
    last_appl. = -1,
    protocols  = 2/10/4 (gcs/repl/appl),
    group UUID = 605a037b-6d84-11e8-a172-8ff5dbd1c9b3
2023-05-31 15:21:44.101 FATAL: /home/buildbot/buildbot/build/gcs/src/gcs_group.cpp:group_check_proto_ver():258: Group requested gcs_proto_ver: 2, max supported by this node: 0.Upgrade the node before joining this group.Need to abort.
2023-05-31 15:21:44.101  INFO: /usr/bin/garbd: Terminated.
2023-05-31 15:23:15.094  INFO: CRC-32C: using 64-bit x86 acceleration.

Although the fatal error could be more helpful, it gives a hint about a protocol version mismatch.

By looking at the currently installed packages on the Arbitrator server, we can still find the older version galera-3:

root@arbitrator:~# dpkg -l|grep galera
ii  galera-3              25.3.37-bionic     amd64        Replication framework for transactional applications
ii  galera-arbitrator-3   25.3.37-bionic     amd64        Galera arbitrator daemon

However, as mentioned above, the data nodes were upgraded to use the newer galera-4. Although the Arbitrator server is not a data node and does not actually sync data, the same version must be used as other nodes in the cluster.

Installation of galera-4:

root@arbitrator:~# apt-get install galera-4 galera-arbitrator-4

Followed by a restart (using garbd process launched by /etc/rc.local):

root@arbitrator:~# cat /etc/rc.local
#!/bin/bash
/usr/bin/garbd --cfg /etc/garbd.conf &
exit 0

root@arbitrator:~# systemctl restart rc-local

And the garbd process started and synced successfully with the Galera cluster:

root@arbitrator:~# tail -f /var/log/garbd.log
2023-05-31 15:25:46.282  INFO: Sending state transfer request: 'trivial', size: 7
2023-05-31 15:25:46.285  INFO: Member 2.0 (garb) requested state transfer from '*any*'. Selected 0.0 (mysql01)(SYNCED) as donor.
2023-05-31 15:25:46.285  INFO: Shifting PRIMARY -> JOINER (TO: 149331740)
2023-05-31 15:25:46.287  INFO: 2.0 (garb): State transfer from 0.0 (mysql01) complete.
2023-05-31 15:25:46.287  INFO: Shifting JOINER -> JOINED (TO: 149331740)
2023-05-31 15:25:46.288  INFO: Member 2.0 (garb) synced with group.
2023-05-31 15:25:46.288  INFO: Shifting JOINED -> SYNCED (TO: 149331740)
2023-05-31 15:25:46.302  INFO: 0.0 (mysql01): State transfer to 2.0 (garb) complete.
2023-05-31 15:25:46.304  INFO: Member 0.0 (mysql01) synced with group.
2023-05-31 15:25:49.281  INFO: (92928494, 'tcp://0.0.0.0:4567') turning message relay requesting off

Add a comment

Show form to leave a comment

Comments (newest first)

No comments yet.