stability: determine reason behind remaining Raft snapshots #8639
Description
#8636 eliminated the majority of Raft initiated snapshots, but a 10m block_writer run against a local 4 node cluster shows about 1/3 of the remaining snapshots are due to Raft:
~/Development/go/src/github.com/cockroachdb/cockroach pmattis/raft-log-queue egrep -h -o 'applied (preemptive|Raft) snapshot' cockroach-data/*/logs/0.stderr | sort | uniq -c
12 applied Raft snapshot
25 applied preemptive snapshot
Naively I would expect 0 Raft snapshots in this test setup.
Metadata
Metadata
Assignees
Labels
No labels