Description
Currently our nemesis logic relies on attempting to track the current state of the cluster in a monolithic stateful nemesis. This makes it difficult to modify the nemesis behaviour without causing unforeseen issues. As a result, some workloads are broken and difficult to fix. Split the nemesis into individual functions for each operation that do not share state to avoid such issues.
Attachments
Issue Links
- has to be done before
-
MB-35293 [Workload] Jepsen partition-failover workload does not function as intended
- Closed
For Gerrit Dashboard: MB-35752 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
114061,2 | MB-35752 Avoid :refer :all | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114062,2 | MB-35752 Cluster run fixes | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114063,2 | MB-35752 Send rebalance rest-call to current control host | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114064,2 | MB-35752 Add util function to query state from ns_server | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114065,3 | MB-35752 Split out nemesis failover logic | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114066,3 | MB-35752 Split out nemesis partition logic | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114070,2 | Revert "MB-35752 Avoid :refer :all" due to it breaking the build | master | jepsen.couchbase | Status: ABANDONED | +2 | -1 |
114073,2 | MB-35752 Fix recovery bug | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114080,3 | MB-35752 Split out nemesis rebalance logic | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114083,2 | MB-35752 Split out nemesis kill logic | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114084,3 | MB-35752 Split out remaining nemesis logic | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
114085,3 | MB-35752 Remove nemesis node state tracking | master | jepsen.couchbase | Status: MERGED | +2 | +1 |