Commits on Jul 2, 2015
  1. Merge pull request #2 from grammarly/master

    Join the cluster if there are other candidate nodes
    committed Jul 2, 2015
Commits on Oct 20, 2014
  1. Merge pull request #3 from lucafavatella/fix-elected-callback-type-spec

    Small type spec improvement in elected callback
    committed Oct 20, 2014
Commits on Oct 2, 2014
  1. @lucafavatella

    Correct type spec of elected callback...

    ... as third argument is `undefined` for leader when elected.
    lucafavatella committed Oct 2, 2014
Commits on Aug 27, 2014
  1. @ybogdanov
Commits on Nov 15, 2013
Commits on Jan 7, 2013
Commits on Nov 19, 2012
  1. Fix dialyzer errors

    committed Nov 19, 2012
Commits on Sep 26, 2012
Commits on Jun 20, 2012
  1. Additional debug printing

    committed Jun 20, 2012
Commits on Mar 2, 2012
Commits on Jan 10, 2012
  1. Increase timeout for setting a monitor on a node

    If a node is down/unreachable, erlang can take longer than 5 seconds to set the monitor
    committed Jan 10, 2012
Commits on Nov 17, 2011
  1. Adding some source comments

    committed Nov 17, 2011
Commits on Aug 18, 2011
  1. @Vagabond

    Break a deadlock when starting a candidate

    When a candidate starts but the leader is behind it in the candidate
    list but all nodes before it are up, it would sit in safe_loop forever.
    This patch implements a fix to the issue.
    Vagabond committed Aug 3, 2011
  2. @Vagabond

    Have new leaders send checklead msgs; Don't omit up nodes from checkl…

    This resolves an issue where checklead messages weren't actually being
    sent because erlang saw the node as up but gen_leader didn't.
    Additionally, the leader elected on the other side of the split wasn't
    sending checklead messages, this has also been resolved.
    Vagabond committed Aug 1, 2011
  3. @Vagabond

    Speed up elections during a netsplit by skipping delayed candidate_ti…

    …mer events
    In a netsplit scenario, sending heartbeats to downed nodes can be very
    expensive (have to wait for TCP timeouts). This causes candidate_worker
    messages to stack up in the mailbox. This patch flushes them to prevent
    that from happening.
    Vagabond committed Jul 26, 2011
  4. @Vagabond

    Trigger a leader election if there's 2 leaders after a netsplit

    When a candidate node goes down, the leader starts sending 'checklead'
    messages to all the downed nodes. If a netsplit occurs and is healed,
    this message is used to tell the cluster that there are competing
    leaders and a new election needs to be held. Credit to Micah Warren
    @lordnull for the original patch.
    Vagabond committed Jul 26, 2011
  5. @Vagabond
Commits on May 23, 2011
  1. @archaelus

    Ignore emacs swap files.

    archaelus committed May 23, 2011
Commits on May 16, 2011
  1. @archaelus
Commits on May 10, 2011
  1. @archaelus
Commits on May 9, 2011
  1. @archaelus
  2. @archaelus
  3. @archaelus
Commits on Apr 7, 2011
Commits on Mar 22, 2011
Commits on Feb 3, 2011
  1. @sg2342
Commits on Jan 28, 2011