diego/0.1481.0

Changes from v0.1480.0 to v0.1481.0

• Verified with garden-linux-release v0.339.0.
• Verified with garden-runc-release v0.4.0.
• Verified with etcd-release v60.
• Verified with cflinuxfs2-rootfs-release v1.20.0.

Significant changes

IMPORTANT: Diego no longer contains a separate ‘converger’ component. Instead, the active BBS server runs Task and LRP convergence periodically and in reaction to a cell presence disappearing.

BBS Relational Datastore

• As a Diego operator, I expect to be able to upgrade my MySQL-backed deployment from the earliest supported version to the latest without downtime

Performance

• As a Diego developer, I would like to have tooling to extract performance metrics from BBS and auctioneer logs

Component Coordination

• As a Diego operator, I expect the active BBS to handle LRP and Task convergence so that I can remove the converger job

SSH

• As a Diego operator, I expect that the diego-sshd process command line does not contain credentials so that unauthorized actors on the Cell VM cannot enter the container via SSH
• cloudfoundry/diego-ssh #22: Rename ‘appplication’ to ‘application’

Routing

• As an app developer, I expect that the system should unregister HTTP routes for instances that are no longer desired
• Investigate Evacuating Instances how we emit the routes in route-emitter.

Garden-RunC Integration

• Look into inigo-etcd-runc failures

Manifest Generation

• CF-MySQL manifest from generate-mysql-bosh-lite-manifest script does not deploy with latest CF-MySQL release-candidate

App Logging and Metrics

• As a CF developer, if bits fail to stream into the container, I expect to see relevant information in the app logs and the crash event so that I can diagnose the failure
• As a CF developer, I expect the Diego healthcheck to log and to exit with different status codes for different types of failures

Dependencies

• As an OSS developer, I expect diego-release to document how to install v0.6.4 of consul so that my local dependencies can match those in consul-release

Test Suites and Tooling

• Investigate “container expired” failures on CI environments
• fix dockerapplifecycle flaky tests

Documentation

• As a Diego operator, I expect to follow documentation about any deployment-order constraints so that I can be sure that my deployment is upgraded safely
• Document the internal Tasks API on the BBS
• Document the internal LRP API on the BBS
• Document the public events API and related types on the BBS

BOSH job changes

Removed converger job.

BOSH property changes

• Added convergence properties to BBS:
  • diego.bbs.convergence.repeat_interval_in_seconds: Interval between runs of the convergence process.
  • diego.bbs.convergence.kick_task_duration_in_seconds: Interval between kicks to tasks.
  • diego.bbs.convergence.expire_completed_task_duration_in_seconds: Duration after which completed, unresolved tasks are deleted.
  • diego.bbs.convergence.expire_pending_task_duration_in_seconds: Duration after which unclaimed tasks are marked as failed.
• Removed converger properties:
  • diego.converger.debug_addr
  • diego.converger.repeat_interval_in_seconds
  • diego.converger.kick_task_duration_in_seconds
  • diego.converger.expire_completed_task_duration_in_seconds
  • diego.converger.expire_pending_task_duration_in_seconds
  • diego.converger.log_level
  • diego.converger.dropsonde_port
  • diego.converger.bbs.api_location
  • diego.converger.bbs.ca_cert
  • diego.converger.bbs.client_cert
  • diego.converger.bbs.client_key
  • diego.converger.bbs.require_ssl
  • diego.converger.bbs.client_session_cache_size
  • diego.converger.bbs.max_idle_conns_per_host

- name: "diego"
  version: "0.1481.0"
  url: "https://bosh.io/d/github.com/cloudfoundry/diego-release?v=0.1481.0"
  sha1: "d3b4b73cd86432c1ec644f3fbe6eb51f3ffc5407"

bosh upload-release --sha1 d3b4b73cd86432c1ec644f3fbe6eb51f3ffc5407 \
  "https://bosh.io/d/github.com/cloudfoundry/diego-release?v=0.1481.0"