Commits on Jun 24, 2015
  1. @nelhage

    Merge pull request #48 from stripe/nelhage-timeout

    Add a --signal-timeout option to escalate to SIGKILL.
    nelhage committed Jun 23, 2015
  2. @nelhage

    Add a --signal-timeout option to escalate to SIGKILL.

    For many applications, it's appropriate to give workers the option to
    gracefully exit, but to time out and escalate to SIGKILL if they
    don't. This prevents misbehaving workers from blocking an upgrade or a
    shutdown indefinitely.
    nelhage committed Jun 23, 2015
Commits on Sep 23, 2014
  1. @zenazn

    Version 0.6.3

    zenazn committed Sep 23, 2014
  2. @zenazn

    Merge pull request #41 from stripe/carl-writeable

    Fix UNIXSocket leak
    zenazn committed Sep 23, 2014
  3. @zenazn

    Fix UNIXSocket leak

    Due to a typo in the event loop code, references to writer IOs would be
    kept around forever in the writer set, which prevented them from being
    GC'd.
    zenazn committed Sep 23, 2014
Commits on Sep 9, 2014
  1. @ebroder

    Bump version to 0.6.2

    ebroder committed Sep 9, 2014
  2. @ebroder
  3. @ebroder
  4. @ebroder
Commits on Aug 22, 2014
  1. @ebroder

    Bump version to 0.6.1

    ebroder committed Aug 21, 2014
  2. @ebroder
Commits on Aug 19, 2014
  1. @ebroder

    And 2.0 has its own serialization format too

    I'm really unclear how they manage to keep coming up with variants
    here.
    ebroder committed Aug 18, 2014
  2. @ebroder
  3. @ebroder
  4. @ebroder

    Add ruby 2.0 fix to bundler

    ebroder committed Aug 18, 2014
  5. @ebroder

    Start testing on 2.0 and 2.1

    ebroder committed Aug 18, 2014
  6. @ebroder
  7. @ebroder
Commits on Aug 6, 2014
  1. @ebroder

    Bump version to 0.6.0

    ebroder committed Aug 5, 2014
  2. @ebroder
  3. @ebroder

    Add an explanatory comment

    ebroder committed Aug 5, 2014
  4. @ebroder
  5. @ebroder
  6. @ebroder
  7. @ebroder

    Add an abstraction for modifying state when dumping

    And use it so that we continue to populate the old
    :reloading_for_preload_upgrade field for downgrades.
    
    (This is mostly useful for downgrade compatibility)
    ebroder committed Aug 5, 2014
  8. @ebroder
  9. @ebroder
Commits on May 28, 2014
  1. @antifuchs

    Bump version to 0.5.8

    antifuchs committed May 27, 2014
  2. @antifuchs

    Merge pull request #37 from antifuchs/clear-env-more-aggressively-on-…

    …reloading
    
    Add switches for altering/restoring the environment on reload
    
    This adds --drop-env-var and --reexec-as, which lets a reloaded einhorn load up as if it had a fresh environment. Also, this changes `upgrade` and `upgrade_fleet` to always reload einhorn, not just on upgrades with preloaded code.
    antifuchs committed May 27, 2014
Commits on May 27, 2014
  1. @antifuchs

    Extract the sentinel process running into its own method

    Also, extract the chunk of code where we set the new master process's
    environment, for OAOOification.
    antifuchs committed May 27, 2014
Commits on May 26, 2014
  1. @antifuchs

    Adjust reload{,ing}_for_preload_upgrade names for the new reality

    It's just reload_for_upgrade now that we always reload einhorn on
    upgrades (:
    antifuchs committed May 26, 2014
  2. @antifuchs
  3. @antifuchs
  4. @antifuchs
  5. @antifuchs