> On 2012-05-09 02:30:29, John Sirois wrote: > > src/slave/slave.cpp, line 1521 > > <https://reviews.apache.org/r/5072/diff/1/?file=107945#file107945line1521> > > > > Was message exchange discussed, pehaps the master periodically > > informing each registered slave of the live executors? This long mtime > > approach seems like ti will work, but it also seems arbitrary.
so this strategy will likely change when we have checkpointing implemented for slave (slave restart stuff) and master (ben's latest state abstraction). for now, this is based on the assumption that no executors are alive when slave restarts. so it is safe to cleanup all executors' directories. > On 2012-05-09 02:30:29, John Sirois wrote: > > src/slave/slave.cpp, line 400 > > <https://reviews.apache.org/r/5072/diff/1/?file=107945#file107945line400> > > > > This path seems like a red flag - it must be repeated elsewhere - can > > you factor up a constant or punch through a config? done. - Vinod ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/5072/#review7715 ----------------------------------------------------------- On 2012-05-09 02:10:27, Vinod Kone wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/5072/ > ----------------------------------------------------------- > > (Updated 2012-05-09 02:10:27) > > > Review request for mesos, Benjamin Hindman, John Sirois, and Brian Wickman. > > > Summary > ------- > > This is the first cut for GC inside the slave. > > There are 2 kinds of gc going on > > --> Executor work dirs -- These get deleted whenever (after a timeout) an > executor exits/shutdown > --> Old slave dirs -- These get deleted when the slave gets registered for > the first time on a startup > > > Diffs > ----- > > src/slave/slave.hpp 08a29d8 > src/slave/slave.cpp 09a8396 > src/common/utils.hpp 1d81e21 > src/slave/constants.hpp f0c8679 > > Diff: https://reviews.apache.org/r/5072/diff > > > Testing > ------- > > make check scucceeds. > > Yet to write GC specific tests. > > > Thanks, > > Vinod > >
