RSS

This Week in Flynn

This week we continued to fix bugs and improve Flynn’s stability.

§ Changes

§ Enhancements

  • Prevent scheduler from scaling down critical apps (#2615)
  • Improve router request error logging (#2643)
  • Update pgx (#2648)
  • Update raft (#2652)
  • Use monotonic time for timeouts (#2655)

§ Bugfixes

  • Fix typo in installer (#2641)
  • Start blobstore during backup restore (#2647)
  • Always include auth token when configured for Github Enterprise (#2657)
  • Use automatic leader election for Redis (#2632)
  • Don’t dereference nil database status (#2656)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

Flynn’s dashboard now has a resource provisioner, which allows users to add database resources to a running app, or to provision a standalone database which can be used by both Flynn apps and external apps.

We’ve also shipped a number of stability improvements.

§ Changes

§ Enhancements

  • dashboard: Add UI for provisioning resources (#2490)

§ Bugfixes

  • installer: Print more readable startup errors (#2637)
  • controller: Don’t require auth for GET /ca-cert (#2635)
  • test/rootfs: Use upstart stop directly (#2630)
  • bootstrap: Support running a subset of steps (#2628)
  • Godeps: Update github.com/thoj/go-ircevent (#2626)
  • updater: Support MariaDB and Sirenia upgrade properly (#2581)
  • Godeps: Update digitalocean client (#2623)
  • discoverd/server: Fix regression sending heartbeats through raft (#2618)
  • controller/worker/deployment: Use configured deploy timeout when waiting for events (#2614)
  • test: Update the CI build log page (#2607)
  • installer: AWS installer: Check all available keys (#2609)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

Flynn now supports process prioritization using the flynn limit command. The priority setting is relative, e.g. when a system is under load, a process with priority 2000 will get double the CPU time of a process with priority 1000. The default priority is 1000.

# Give the web process double the default priority of 1000
flynn limit set web cpu=2000

# Give the worker process a lower priority
flynn limit set worker cpu=500

We’ve also continued work on stability.

§ Changes

§ Enhancements

  • Add CPU limits and resource partitions (#2590)

§ Bugfixes

  • dashboard: Use proper deploy timeouts (#2605)
  • updater: Fix updating slug environment variables (#2602)
  • test/rootfs: Add MongoDB unit test dependencies (#2565)
  • pkg/postgres: Fix connection leak when Listener errs out (#2599)
  • Fix cluster backup bugs (#2598)
  • host: Add 64MB /dev/shm to each container (#2596)
  • host/cli: Use current MariaDB artifact URI when restoring from backup (#2592)
  • discoverd/server: Fix mutex that doesn’t get unlocked in error path (#2591)
  • Fix host-wait timeout (#2589)
  • cli: Fix backup/export progress info (#2568)
  • installer: Use –clean when using install script (#2567)
  • host/cli: Send stderr from log stream to stdout by default (#2569)
  • Fix formation stream leak (#2574)
  • util/release: Add MariaDB to release manifest (#2571)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

Flynn now has a MySQL database appliance available in beta, to go along with Postgres and Redis. Flynn’s MySQL appliance features the same highly-available configuration as our Postgres appliance, and supports app and cluster backups. Run flynn resource add mysql to provision a new MySQL database for your app.

A number of bugfixes also landed, as we focus on stability and performance.

We’re starting to see a number of Flynn-specific job posting in the wild. Most recently we were asked by Vestorly, a current Flynn user, to advertise their post for a DevOps engineer who is passionate about automation and owning an environment. If you are interested, visit their careers page.

§ Changes

§ Enhancements

  • Mariadb integration (#2433)

§ Bugfixes

  • dashboard: Fix scaling app processes when no formation (#2566)
  • dashboard: Remove stale files (#2564)
  • Godeps: Bump asset-matrix-go (#2563)
  • Pre-Go1.6 fixes (#2562)
  • appliance/mariadb: Use more restrictive quoting (#2560)
  • Fix containerinit registeration on error (#2557)
  • status: Optional services no longer mark cluster unhealthy (#2554)
  • pkg/httpclient: Don’t use a timeout in ResumingStream (#2552)
  • scheduler: Don’t try to start jobs which have already started (#2551)
  • slugbuilder: Bump buildpacks (#2549)
  • Add logging to containerinit health monitor (#2543)
  • scheduler: Ensure omni jobs do not become unbalanced (#2542)
  • worker: Watch job events per release when deploying (#2537)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This week we fixed a variety of bugs in Flynn, improving stability.

§ Changes

§ Improvements

  • Improve CI management scripts (#2506, #2510)
  • Add leader flag to routes (#2395)
  • Raise postgres max connections to 400 (#2519)
  • Bump CI test timeout to 35 minutes (#2530)
  • Make max concurrent add job calls configurable (#2529, #2531)

§ Bugfixes

  • Add job add ratelimit to prevent overload during backup restore (#2503)
  • Fix containerinit hang (#2505)
  • Fix panic during install when there is no SSH agent (#2518)
  • Fix deadlock during state restoration (#2522)
  • Fix hard-coded high port in router test (#2515)
  • Bump discoverd leader wait at startup (#2512)
  • Update local cluster config after domain migration (#2516)
  • Log domain migration stream error (#2533)

§ Stay in Touch

§ How You Can Help

 RSS



Mailing List