In a cluster of three broker nodes configured in full-sync mode. For the first time when the leader node becomes unavailable, the follower node becomes leader in less time as compared to when the second time the leader node becomes unavailable.
Is this an expected behavior? for more understanding can we know, which leader election algorithm is used in Full-Sync mode.
Hi,
The times can vary a little bit depending on the timings of the disconnect of the node.
We are using the Consensus Algorithm from Raft.
They have a beautiful site that also visualizes the process.
Depending on what time you disconnect at some time another node expects a message from the leader, if that does not happen, it will start an election process.
Hi Tizian,
Can you please share the standard benchmark values here (considering what system configuration, expected disconnect notification etc).