Concerning these two parameters in Galera configuration, evs.keepalive_period and evs.inactive_check_period
Looking in Galera documentation the default values for both is PT1S
Also, the relation between the two parameters is
evs.keepalive_period <= evs.inactive_check_period
But in percona xtradb documentation, the default values are
evs.keepalive_period = PT1S
evs.inactive_check_period = PT0.5S
Then the above relation is not satisfied .
Please, can you explain why for xtdadb , the defaul vakues are different and the relation is not satisfied ?
Also, il seems that ffor checking peer connectivity, each node in galera cluster uses gmcast.peer_timeout
Can you explain how each node check the reachabilty of the other nodes. I mean which parameter is used ( gmcast.peer_timeout or evs.keepalive_period ) ?
The relation between the two parameters is found here
As you can see, it s reported that
evs-keepalive-period) <= evs.inactive_check_period
Also, for gmcast.peer_timeout , what do you mean by "relaying protocol " ?
There is no documentation describing how it works.
I can guess it is used by Galera to fetch info about one node by contacting other nodes.
Please consider a three-node cluster. In normal circumstances, each node sees all other nodes. Now imagine that communication between nodes 1 and 3 is blocked for any reason. In such a case, Galera activates a low-level protocol that allows node 1 to communicate with node 3 via node 2. This is a relaying protocol.