> On March 17, 2015, 7:49 p.m., Zameer Manji wrote: > > src/main/python/apache/aurora/executor/common/path_detector.py, line 34 > > <https://reviews.apache.org/r/32013/diff/3/?file=896451#file896451line34> > > > > What's the rationale behind this change? I think the callers should be > > doing the checking for existance and dealing with missing directories as > > needed.
The rationale is that some tasks will have checkpointing in the sandbox (0.9.0+) and some will have it outside the sandbox (<0.8.0, or 0.9.0+ but whose thermos runners haven't yet been forked.) A typical production slave has 1000-2000 leftover framework directories, but since we haven't rolled sandbox checkpointing yet, that's 1000-2000 TaskDetectors we must construct and query if we don't filter by tasks that have this feature enabled. That filtering definitely seems like it belongs in the MesosPathDetector. - Brian ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/32013/#review76784 ----------------------------------------------------------- On March 17, 2015, 8:24 p.m., Brian Wickman wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/32013/ > ----------------------------------------------------------- > > (Updated March 17, 2015, 8:24 p.m.) > > > Review request for Aurora, Joe Smith and Zameer Manji. > > > Repository: aurora > > > Description > ------- > > SSIA > > > Diffs > ----- > > src/main/python/apache/aurora/executor/common/path_detector.py > e4135cca09b43e204c52774b747f74212d0a5732 > src/main/python/apache/thermos/observer/detector.py > 5347cded7061b085b135d62fdb3a5c4a75f81e5c > src/test/python/apache/aurora/executor/common/test_path_detector.py > 2d6edae420c1c5b51fe93f97e04c24976afacd51 > > Diff: https://reviews.apache.org/r/32013/diff/ > > > Testing > ------- > > ./pants test src/test/python/apache/thermos/observer:: + manual observation > > > Thanks, > > Brian Wickman > >
