2.6.1 - 2017-01-23

Fixed

• processor: open log writer early, prevent panic

2.6.0 - 2017-01-17

Security

• update to go v1.7.4

Added

• cli: log processor pool total on shutdown
• amqp_job: meter for job finish state (#184)
• amqp_job: track queue time via queued_at field from payload
• amqp_job_queue: log job id immediately after json decode
• capture every requeue and send the error to sentry

Changed

• image/api_selector: selection of candidates by multiple groups
• amqp_canceller: change verbosity of canceller missing job id to debug
• docker: SIGINT STOPSIGNAL for graceful shutdown

Fixed

• processor: always mark job as done when Run finishes
• processor: use errors.Cause when checking error values
• backend/jupiterbrain: parse SSH key on backend init (#206)
• backend/jupiterbrain: add sleep between creating and wait-for-ip
• backend/docker: run bash with -l (login shell) in docker native mode
• image/api_selector: check job-board response status code on image selection
• image/api_selector: check tagsets for trailing commas before querying job-board
• amqp_job_queue: handle context cancellation when delivering build job
• ssh: request a 80x40 PTY

2.5.0 - 2016-10-03

Added

• support for heartbeat URL checks a la legacy
  worker
• max build log length is now configurable
• alpine-based Docker image
• added runtime-configurable script hooks.

Changed

• check flags and env vars for start and stop hooks

Fixed

• Handling "false" as a valid boolean false-y value
• logging for start/stop hooks is clearer now
• logging for jupiter-brain boot timeouts is more verbose
• AMQP connections get cleaned up as part of graceful shutdowns

2.4.0 - 2016-09-08

Added

• backend/cloudbrain
• backend/docker: "native" script upload and execution option
• backend/gce: Support for instances without public IPs
• handle SIGWINCH by gracefully shutting down processor pool and pausing

Changed

• (really) static binaries built on go 1.7.1, cross compiled natively
• step/run-script: Add a new line with a link to the docs explaining
  travis_wait when there's a log timeout error.
• switch to keepachangelog format
• upgrade all dependencies
• backend/jupiterbrain: improved error logging via pkg/errors

Removed

• Using goxc for cross-compilation
• amqp-job: logging channel struct

Fixed

• Handling "false" as a valid boolean false-y value

2.3.1 (May 30, 2016)

IMPROVEMENTS:

• backend/gce: Allow the per-job hard timeout sent from scheduler to be
  longer than the global GCE hard timeout.

BUG FIXES:

• package/rpm: Fix the systemd service file so the Worker can start.

2.3.0 (April 12th, 2016)

FEATURES:

• backend/gce: Add support for separate project ID for images via
  IMAGE_PROJECT_ID env var.

IMPROVEMENTS:

• amqp-job: Add support for custom amqp TLS certificates via --amqp-tls-cert
  and --amqp-tls-cert-path options.

BUG FIXES:

• backend/gce: Requeue jobs on preempted instances (instances preemptively
  shutdown by gce).

2.2.0 (February 1st, 2016)

IMPROVEMENTS:

• backend/gce: switched out rate limiter with Redis-backed rate limiter
• backend/gce: create a pseudo-terminal that's 80 wide by 40 high instead of 40 wide by 80 high
• backend/gce: make using preemptible instances configurable
• build: switch to Go 1.5.2 for Travis CI builds
• cli: add config option (--amqp-insecure) to connect to AMQP without verifying TLS certificates

2.1.0 (December 15th, 2015)

FEATURES:

• backend/gce: Add support for a "premium" VM type.

IMPROVEMENTS:

• cli: Show current status and last job id via SIGUSR1.
• backend/gce: Use multistep
  during instance creation and deletion.
• backend/gce: Allow configuration of script upload and startup timeouts.
• backend/gce: Various improvements to minimize API usage:
  • Exponential backoff while stopping instance.
  • Add option to skip API polling after initial instance delete call.
  • Add metrics to track API polling for instance readiness.
  • Cache the instance IP address as soon as it is available.
  • Add ticker-based rate limiting.
• backend/gce: Generate SSH key pair internally.
• amqp-job: Track metrics for (started - received) duration.
• cli: Add more percentiles to Librato config.
• config: DRY up the relationship between config struct & cli flags.

BUG FIXES:

• backend/jupiterbrain: Ensure image name is present in instance line of
  worker summary.
• backend/docker: Pass resource limitations in HostConfig.
• backend/docker: Prioritize language tags over default image.

2.0.0 (December 9th, 2015)

FEATURES:

• step/open-log-writer: Display revision URL and startup time in worker info
  header.

IMPROVEMENTS:

• Use goxc for cross-compilation, constrained to
  linux amd64 and darwin amd64.
• backend/gce: Tweaks to minimize API calls, including:
  • Add a sleep interval for before first zone operation check.
  • Break out of instance readiness when RUNNING or DONE, depending
    on script upload SSH connectivity for true instance readiness.
  • Increase default script upload retries to 120.
  • Decrease default script upload sleep interval to 1s.
• backend/gce: Remove legacy image selection method and configuration.

BUG FIXES:

• step/run-script: Mark jobs as errored when log limit is exceeded,
  preventing infinite requeue.

1.4.0 (December 3rd, 2015)

FEATURES:

• backend/docker: Allow disabling the CPU and RAM allocations by setting the
  config options to 0 (this was possible in previous versions, but was not
  documented or supported until now)

IMPROVEMENTS:

• backend/gce: Send job ID and repository slug to image selector to help in
  log correlation
• sentry: Only send fatal and panic levels to Sentry, with an option for
  sending errors as well (--sentry-hook-errors)
• image/api_selector: Send os:osx instead of the language when querying for
  an image matching an osx_image flag

BUG FIXES:

• amqp_job: Send correct received_at and started_at timestamps to hub in
  the case of the job finishing before the received or started event is sent