On config changes, the precommit index will be updated instantly to index of config change.
After it, request_append_entries is called which instantly commits up to precommit index if a node was removed from 2-node cluster.
If logs were being applied at the same time it could lead to commit thread trying to commit logs that were still not precommitted.
I couldn't find another explanation for such scenario.
If my conclusions are wrong feel free to correct me. :pray:
I also couldn't find a better way to fix it quickly.
On config changes, the precommit index will be updated instantly to index of config change. After it,
request_append_entriesis called which instantly commits up toprecommitindex if a node was removed from 2-node cluster. If logs were being applied at the same time it could lead to commit thread trying to commit logs that were still not precommitted. I couldn't find another explanation for such scenario. If my conclusions are wrong feel free to correct me. :pray: I also couldn't find a better way to fix it quickly.