Trying to rollback to checkpoint N from empty recovery file <path>

[slfritchie]

Is this a bug, feature request, or feedback?

Bug

What is the current behavior?

Intermittent failure (very roughly 1 failure out of 10 attempts) of make integration-tests-testing-conformance-all debug=true pytest_exp='-k system_events -k resilience_pony' resilience=on, see full logs at http://wallaroolabs-dev.s3.amazonaws.com/logs/logs.1574730559.tar.gz

--- worker1 (pid: 11176, rc: 1) ->

1574728171.407135,DEBUG,TCPSink,Receive barrier CheckpointRollbackBarrierToken(Rollback 1, Checkpoint 6) at TCPSink 281649526051898532588300641051597128013

1574728171.407141,DEBUG,TCPSink,Receive barrier CheckpointRollbackBarrierToken(Rollback 1, Checkpoint 6) at TCPSink 281649526051898532588300641051597128013

1574728171.407147,DEBUG,TCPSink,Barrier CheckpointRollbackBarrierToken(Rollback 1, Checkpoint 6) complete at TCPSink 281649526051898532588300641051597128013

1574728171.408084,|~~ - Recovery Phase: AwaitDataReceiversAck - ~~|
1574728171.408105,|~~ - Recovery Phase: Rollback - ~~| 
1574728171.408132,Trying to rollback to checkpoint 6 from empty recovery file /tmp/res-data.flvsowex/multi_partition_detector-worker1-000000000000EC1B.evlog
This should never happen: failure in /home/vagrant/wallaroo/lib/wallaroo/core/recovery/backend.pony at line 163

--- worker1 (pid: 11176, rc: 1) <-

What is the expected behavior?

No crash

What OS and version of Wallaroo are you using?

Ubuntu 16.04/Xenial LTS + master @ commit da0491f + PR #3071 branch backend-path-error

Steps to reproduce?

make integration-tests-testing-conformance-all debug=true pytest_exp='-k system_events -k resilience_pony' resilience=on