-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/47732/
-----------------------------------------------------------

Review request for mesos and Vinod Kone.


Repository: mesos


Description
-------

There is a possible race condition when the `TASK_FINISHED`
update calls into `update` when the executor process might
itself be already executing. Also after this fix, even if the 
executor is disconnected from the agent by the time the `update`
function is called, the updates would be sent again as part of 
unacknowledged updates on re-registration.


Diffs
-----

  src/examples/long_lived_executor.cpp 94379b01c3ac8dbf513559789046677bceea688c 

Diff: https://reviews.apache.org/r/47732/diff/


Testing
-------

make check

This is a bit hard to reproduce due to the random delay. e.g., this happened 
once in ~2-3 days on our internal cluster.

```
Backtrace:
./long-lived-executor(_ZNSt4listIN2id4UUIDESaIS1_EE5eraseESt14_List_iteratorIS1_E+0x14)[0x41ffec]
./long-lived-executor[0x41cf81]
./long-lived-executor[0x41ad3a]
./long-lived-executor[0x41f245]
./long-lived-executor[0x424622]
./long-lived-executor(_ZN7process11ProcessBase5serveERKNS_5EventE+0x2e)[0x419010]
/opt/mesosphere/lib/libmesos-0.29.0.so(_ZN7process14ProcessManager6resumeEPNS_11ProcessBaseE+0x2d1)[0x7fa3eba5cc21]
/opt/mesosphere/lib/libmesos-0.29.0.so(+0x13b1f27)[0x7fa3eba5cf27]
/lib64/libstdc++.so.6(+0xb5220)[0x7fa3e9f43220]
/lib64/libpthread.so.0(+0x7dc5)[0x7fa3ea19ddc5]
/lib64/libc.so.6(clone+0x6d)[0x7fa3e99ad28d]
```


Thanks,

Anand Mazumdar

Reply via email to