Skip to content
Commits on Jan 9, 2014
  1. @technomancy
  2. @technomancy

    Return modified State from close_if_idle.

    technomancy committed
    Also multiply ms instead of dividing us in order to avoid floats.
    
    Same as 15366ff, but for HTTP drains.
  3. @technomancy
  4. @technomancy
  5. @technomancy
Commits on Jan 7, 2014
  1. @technomancy
  2. @technomancy
  3. @technomancy
  4. @technomancy
Commits on Jan 6, 2014
  1. @technomancy
  2. @technomancy
  3. @technomancy
Commits on Dec 23, 2013
  1. @technomancy
Commits on Dec 21, 2013
  1. @technomancy
Commits on Dec 20, 2013
  1. @technomancy
  2. @technomancy

    Have a single timer for TCP drains that checks last_good_time.

    technomancy committed
    As per Fred's suggestion, this is a lot simpler than resetting the
    timer every time you buffer a new message since it doesn't require
    re-cancelling timers over and over in a non-idle buffer.
Commits on Dec 19, 2013
  1. @technomancy

    Tidy up a few style issues.

    technomancy committed
  2. @technomancy

    Attempt to read from idled socket in test.

    technomancy committed
    This is better than checking the process state, which was the
    workaround we were using for the fact that port_info refused to
    consider the socket closed.
Commits on Dec 18, 2013
  1. @technomancy
  2. @technomancy
Commits on Dec 17, 2013
  1. @technomancy
  2. @technomancy

    Test against drain's internal state rather than checking for closed s…

    technomancy committed
    …ocket.
    
    For some extremely mysterious reason, closing one side of this socket
    refuses to close the other in the tests, though manual tests with
    netcat confirm that it's in fact being closed.
  3. @technomancy

    Re-cache OS env-vars in TCP drain tests.

    technomancy committed
    Otherwise we get stuck with the default idle timeout.
  4. @technomancy

    Add test for tcp drain closing.

    technomancy committed
    Test is currently failing even though according to the logs the socket
    is being closed at the right spot.
Commits on Dec 16, 2013
  1. @technomancy
Commits on Dec 14, 2013
  1. @technomancy

    Add disconnecting state for when we hit the idle timeout in the sendi…

    technomancy committed
    …ng state.
    
    The disconnecting state is similar to sending, but unless a post is
    received it will end up in the disconnected state. I've collapsed the
    two inet_reply clauses into a single one that only logs when there's
    an error.
  2. @technomancy
Commits on Dec 13, 2013
  1. @technomancy
  2. @technomancy
  3. @technomancy
  4. @technomancy
Commits on Dec 10, 2013
  1. @omarkj

    Merge pull request #67 from heroku/redis_buffer_shard_key

    omarkj committed
    Redis buffer shard key
Commits on Dec 4, 2013
  1. @omarkj
  2. @omarkj

    Upgrade script for both up- and downgrades from-

    omarkj committed
    and to v69.12.
  3. @omarkj

    Add rollback code change.

    omarkj committed
Something went wrong with that request. Please try again.