Permalink
Commits on Jul 13, 2016
  1. @raylu

    Version 0.7.4

    raylu committed Jul 13, 2016
  2. @raylu

    Don't use Timeout and don't catch ECHILD

    Timeout is flaky
    
    We have a pid - one of the waitpid2s in subprocess must succeed,
    so we should never get ECHILD
    raylu committed Jul 13, 2016
  3. @raylu

    Re-add :state_passer so we don't log_error in mourn

    :state_passer children were originally in einhorn
    They were removed in 673f5a8 and a18e09b
    The bug that prevented us from reaping them was fixed in dad306b
    
    Before this commit, upgrades would cause the log_error at the top of mourn to run
    raylu committed Jul 13, 2016
Commits on Jul 12, 2016
  1. @raylu
Commits on Jul 11, 2016
  1. @raylu

    Version 0.7.3

    raylu committed Jul 11, 2016
  2. @ebroder
  3. @raylu

    Merge pull request #64 from stripe/raylu-pid_rollover

    Fix PID rollover
    raylu committed on GitHub Jul 11, 2016
Commits on Jul 8, 2016
  1. @ebroder

    Demote client connection log lines to debug

    I can't think of any cases where these log lines are useful, and they can be quite noisy.
    ebroder committed on GitHub Jul 8, 2016
Commits on Jul 7, 2016
  1. @raylu

    Use :spinup_time to track PID rollover

    raylu committed Jul 7, 2016
  2. @raylu

    Fix PID rollover bug

    Also, don't send SIGKILL for signal commands from einhornsh
    raylu committed Jul 7, 2016
  3. @raylu
Commits on Jul 6, 2016
  1. @nelhage

    Merge pull request #63 from stripe/nelhage-setpgrp

    Launch children in a new process group.
    nelhage committed on GitHub Jul 6, 2016
  2. @nelhage

    constrain Rack version to fix Travis

    nelhage committed Jul 6, 2016
  3. @nelhage

    Launch children in a new process group.

    This has the effect that, when einhorn is run interactively, signals
    sent by the terminal (e.g. in response to ^C) only go to the einhorn
    master, and not to any of its children.
    nelhage committed Jul 6, 2016
Commits on Jun 29, 2016
  1. @jessitron

    Merge pull request #62 from jessitron/patch-1

    Update Ruby version and leader
    jessitron committed on GitHub Jun 29, 2016
  2. @jessitron

    Update Ruby version and leader

    We don't support 1.8.7 anymore, and afaict it's 2.0-2.2 instead. Advertise this.
    
    Also, Greg doesn't work here anymore. Carl has last commit!
    
    r? @zenazn
    jessitron committed on GitHub Jun 29, 2016
Commits on Jun 18, 2016
  1. @zenazn

    Cut version 0.7.2

    zenazn committed Jun 18, 2016
  2. @zenazn

    Merge pull request #61 from stripe/carl-exponential-backoff

    Cap exponential backoff by a reasonable max
    zenazn committed on GitHub Jun 18, 2016
Commits on Jun 17, 2016
  1. @zenazn

    Deprecate Ruby 1.9.x, ree

    zenazn committed Jun 17, 2016
  2. @zenazn

    Cap exponential backoff by a reasonable max

    zenazn committed Jun 17, 2016
Commits on Feb 17, 2016
  1. @zenazn

    Merge pull request #57 from stripe/carl-one-thousand

    Allow counts up to 1,000
    zenazn committed Feb 17, 2016
  2. @zenazn

    Update Travis config

    zenazn committed Feb 17, 2016
  3. @zenazn

    Allow counts up to 1,000

    zenazn committed Feb 17, 2016
Commits on Oct 27, 2015
  1. @nelhage

    Einhorn 0.7.1

    nelhage committed Oct 27, 2015
  2. @nelhage

    Merge pull request #53 from stripe/nelhage-retry

    Don't resend commands on EPIPE.
    nelhage committed Oct 27, 2015
  3. @nelhage

    Don't resend commands on EPIPE.

    We could try to add back in a retry loop at a higher level, but it's not
    worth it. This behavior is clearly wrong.
    nelhage committed Oct 27, 2015
  4. @nelhage

    Merge pull request #52 from stripe/nelhage-state

    `state` command: Don't dump descriptors.
    nelhage committed Oct 27, 2015
  5. @nelhage

    Maintain 1.8 compatibility.

    Because I can't be bothered to delurk Ruby 1.8 right now.
    nelhage committed Oct 27, 2015
  6. @nelhage

    `state` command: Don't dump descriptors.

    These are an internal implementation detail and not unlikely to be
    useful to clients (we currently have no users), and having this in the
    `state` output risks an exponential explosion in the size of the
    serialized state, in the event of multiple concurrent `state`
    commands (since the descriptor representing the `einhornsh` client will
    contain a complete copy of the YAML-encoded previous state).
    nelhage committed Oct 27, 2015
Commits on Aug 20, 2015
  1. @antifuchs

    Version 0.7.0

    antifuchs committed Aug 20, 2015
  2. @antifuchs

    Merge pull request #50 from stripe/asf-remove-plugins

    Remove plugin support
    antifuchs committed Aug 20, 2015
  3. @antifuchs

    Remove plugin support

    This change removes little_plugger, and the entire plugin facility from
    Einhorn. The functionality hasn't really seen any use. In addition,
    little_plugger has the nasty habit of rescanning all directories on $:
    for new plugins, which in combination with many installed gems slows
    down restarts a lot if the machine's VFS cache is cold.
    antifuchs committed Aug 20, 2015
Commits on Jul 31, 2015
  1. @zenazn

    Bump version to 0.6.5

    zenazn committed Jul 31, 2015
  2. @zenazn

    Flush loglines atomically

    Ruby's IO#puts calls write(2) twice: once to write the string, and once
    to write a newline. This means that other writes to the same file (in
    other processes, for instance) might be interleaved between einhorn log
    lines. If lines explicitly end in the newline character, however, Ruby
    will omit its newline and thus the second call to write(2).
    zenazn committed Jul 31, 2015
Commits on Jun 24, 2015
  1. @nelhage

    Version 0.6.4

    nelhage committed Jun 24, 2015