v1.4.0

• [github] #91: Support for mounting host directories into VMs. Details can be found at https://github.com/veertuinc/anklet#host-to-guest-folder-mounts
• [github] Both handlers are immediately paused when API limits are reached. Fixed bug where it was retrying the job over and over instead of just pausing until the api limits are reset.

v1.3.0

Allow overcommitting of resources

In this release we're introducing global_skip_cpu_and_memory_resource_checks and skip_cpu_and_memory_resource_checks options, allowing VMs to schedule no matter their resources. By default, Anklet will not start a VM on a host until there are enough resources. This works well, but can slow down queue times. Especially when the VMs may be set to allow for X CPUs and RAM, but the workloads inside would never use more than half the available host resources.

Developer note: Plugins will need to ensure support for this. Right now it's just for github, but in the future all plugins must suppor this feature.

Globally

---
work_dir: /tmp/
pid_file_dir: /tmp/
. . .
global_skip_cpu_and_memory_resource_checks: true # Optional; skip VM CPU/RAM admission checks for all handler plugins (allows overcommit).
plugins:
. . .

Plugin level

plugins:
  - name: RUNNER1
    plugin: github
    . . .
    skip_cpu_and_memory_resource_checks: true # Optional; allows CPU/RAM overcommit by skipping VM admission checks on this host.

v1.2.2

• [github receiver] Fix receiver crash when unmarshaling repository owner from GitHub webhook payloads (on redelivery)

v1.2.1

• Preventing "paused" from becoming stuck in the metrics endpoint.
• Various nil pointer prevention bug fixes.

v1.2.0

• Ability to use ENVs to override (or set) individual plugin configs

v1.1.1

• [github receiver] moved in_process cleanup AFTER redelivery
• [github receiver] fix for sleep when redelivering

v1.1.0

• [github receiver & handler] Add queue_name config so that receivers and handlers can target the same queue, no matter the org registered
• [github handler] Ensure handlers only run jobs that match their org.

v1.0.1

• Security patches

v1.0.0

• Test suite.
• [github] Bug fix: Two jobs start ~100ms apart. Second job gets scheduled, even though it should be paused because the first job would take up too much. However, it was not happening properly.
• [github] Secondary rate limit tracking and pausing.
• [github] Don't increment attempts if it's rejected on a host because of a pull running.
• [github] timeout executing register-runner.bash would crash anklet
• Handle SIGINT properly

v0.19.0

• github | receiver: Prevent duplicate events from being queued by checking handler queues as well as the main one
• github | receiver: Fixed redelivery problem where nothing was being redelivered and also only the first 10 items were being attempted