This is an automatically generated e-mail. To reply, visit:

(Updated March 23, 2016, 3:14 p.m.)

Review request for mesos, Adam B, Alexander Rojas, and Neil Conway.


Review comment from Adam

Bugs: MESOS-4984

Repository: mesos

Description (updated)

We were not correctly waiting for the master to register the first
slave before making a call to the `/slaves` endpoint. There was this
possible race:

- Slave1 is started.
- Slave2 is started.
- Slave2 sends register message to master.
- Slave2 re-tries the register message.
- Master registers slave2.
- Master resends register acknowledgment to slave2.
- The test thinks that both expectations i.e. `FUTURE_PROTOBUF`
have completed.
- Test makes call to `/slaves` endpoint and sees 1 slave has only registered. 
The test fails.
- Master registers slave1.

Reordering the `AWAIT_READY` calls to explicitly wait for
`slave1` to register first should fix the flakiness.

Diffs (updated)

  src/tests/master_tests.cpp d34ba0bdd71efd261850d8c205c16cecb701ac7c 

Diff: https://reviews.apache.org/r/45166/diff/


Ran it in a loop. Previously used to fail quite often.


Anand Mazumdar

Reply via email to