RSS

This Week in Flynn

This week we shipped many bugfixes and continued to improve Flynn’s stability.

§ Changes

§ Enhancements

  • Combine Entrypoint and Cmd into one Args field (#3138, #3147, #3149)
  • Update pgx (#3140)
  • Show Raft status in flynn-host list (#3131)
  • Add development script for installer (#3128)
  • Improve mariadb integration test logging (#3114)
  • Use latest kernel from Ubuntu Xenial (#3121)
  • Add pkg-config to CI rootfs (#3117)
  • Add squashfs-tools to CI rootfs (#3116)
  • Don’t start cluster with MongoDB running (#3103)

§ Bugfixes

  • Install libvirt when installing older versions (#3099)
  • Simplify Upstart start on triggers (#3139)
  • Stop sirenia peer before heartbeater (#3135,
  • Fix host crash due to race and subsequent recovery (#3133)
  • Retry service registration and heartbeater creation (#3122)
  • Don’t install libvirt during TestReleaseImages (#3119)
  • Prevent empty environment variable keys (#3112)
  • Start scheduler persistence loop before syncing jobs (#3110)
  • Fix mariadb waitReadWrite in integration tests (#3109)
  • Return validation error when killing pending job (#3101)
  • Remove unnecessary kernel capabilities (#3096)
  • Mark unknown pending jobs as down (#3098)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This week we changed our underlying container runner from libvirt-lxc to libcontainer, which improves stability, fixes several bugs, and halves container start time. Additionally, Postgres was updated from 9.4 to 9.5, the job crash restart algorithm was improved to react much faster, and many other bugs were fixed.

§ Changes

§ Enhancements

  • Upgrade to Postgres 9.5 (#2952)
  • Improve distribution release system (#3062, #3068, #3071)
  • Improve sirenia resurrection and recovery (#3064)
  • Show error when pushing to non-master branch (#3089)
  • Replace libvirt-lxc with libcontainer (#3030)
  • Simplify crash restart backoff algorithm (#3082)
  • Enable TCP keep-alive (#3079)

§ Bugfixes

  • Install test dependencies in Vagrant development VM (#3046, #3086)
  • Fix version in TestReleaseImages (#3074)
  • Add CA root certificates to docker-receive image (#3069)
  • Fix status page when cluster is unhealthy (#3066)
  • Fix parsing of double space in blobstore backend config (#3056)
  • Remove broken link on development page (#3058)
  • Bump CI rootfs to 50G (#3057)
  • Only start Docker in development VM (#3055)
  • Fix router TLS migration bugs (#3049)
  • Fix dashboard links to TLS and path-based routes (#3051)
  • Synchronize scheduler test when changing host tags (#3093)
  • Remove extraneous mutex lock in router (#3088)
  • Fix slugignore directory exclusion (#3090)
  • Fix failed attach permanently blocking (#3091)
  • Fix flannel backend variable in development script (#3084)
  • Remove sirenia test deploy timeout (#3085)
  • Don’t rollback one-by-one deploy if cleanup fails (#3080)
  • Log formation expansion error in controller stream (#3078)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This week we added a MongoDB appliance to Flynn, and continued to fix bugs and improve stability.

§ Changes

§ Enhancements

  • Add HTTP endpoint to scheduler that dumps internal state (#2998)
  • Add --version and --channel flags to install script (#3003)
  • Use internal state in scheduler tests (#3011)
  • Add development script for dashboard development (#3024)
  • Add flynn-host discover command (#2166)
  • Add libseccomp-dev to CI rootfs (#3033)
  • Use default flynn-host backend in CI (#3034)
  • Add MongoDB appliance (#2561, #3039, #3038, #3041, #3044)
  • Add link to show system apps in dashboard (#3032)

§ Bugfixes

  • Delete releases when deleting apps (#3007)
  • Follow redirects when downloading slugs (#3016)
  • Make controller shutdown quickly and cleanly (#3014)
  • Fix router domain migration (#3013, #3027)
  • Fix race when closing postgres listener (#3025)
  • Respect app deploy timeout in sirenia integration tests (#3042)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This week we shipped some new docs, including MySQL, Redis, and Docker.

§ Changes

§ Enhancements

  • Update Docker deploy docs (#2994)
  • Update database and operating docs (#2992)
  • Add docs to development script (#2996)
  • Add support for SSH key auth to taffy (#2932)
  • Add libsasl2 to rootfs (#2967)

§ Bugfixes

  • Unblock AddJob ratelimit while waiting for bootstrap (#2995)
  • Fix various things in CLI help (#2991)
  • Fix govet complaints (#2950)
  • Fix controller hangs during postgres failover (#2953)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This week many significant improvements and bugfixes landed. Some highlights are that we’ve finally updated to Go 1.6, the blobstore gained support for S3, the router now supports HTTP/2, and the CLI has two new commands: flynn release rollback and flynn deployment timeout. Check out the list below for the details.

§ Changes

§ Enhancements

  • Update redis to v3.0.7 (#2957)
  • Wait for web process to start successfully on first push (#2958)
  • Add flynn release rollback command (#2954)
  • Cancel deploy when git push is terminated (#2951)
  • Cancel backend HTTP requests when client disconnects from router (#2951)
  • Add support for HTTP/2 to router (#2948, #2955)
  • Add support for S3 to blobstore (#2922, #2942)
  • Use vendor directory instead of Godeps workspace (#2947)
  • Add cluster status page to dashboard (#2930, #2928)
  • Update to Go 1.6 (#2558)
  • Update ubuntu rootfs URL (#2944)
  • Increase CI VM memory allocation (#2943)
  • Remove controller deploy stream timeout (#2934)
  • Add flynn deployment timeout command (#2933)
  • Make discoverd promote/demote idempotent (#2919)
  • Only log failed requests to discoverd (#2914)
  • Update Raft package (#2931)
  • Refactor TLS certificate handling in router (#2890, #2962)
  • Improve log view in dashboard (#2907)

§ Bugfixes

  • Fix scheduler test races (#2961)
  • Fix logaggregator test race (#2959)
  • Fix Docker image import tagging (#2956)
  • Enable router upgrade path test (#2949)
  • Guard discoverd against decoding nil deploy metadata (#2945)
  • Fix DNS edge case during discoverd promote/demote (#2946)
  • Fix flaky discoverd cluster integration test (#2940)
  • Fix various races (#2558)
  • Reset HostID when scheduler is placing jobs (#2939)
  • Fix deleting apps that are scaled down (#2920)
  • Tolerate missing docker-receive status (#2925)

§ Stay in Touch

§ How You Can Help

 RSS



Mailing List