RSS

Flynn Stable Channel and San Francisco Meetup

We’re excited to announce the launch of the stable release channel for Flynn today.

We will be hosting a Flynn meetup at the AWS Loft in San Francisco next Tuesday, November 10th from 7-8:30pm. RSVP here.

Stability

Our highest priority is making Flynn as reliable as possible in production environments. This release channel and the current version of Flynn are both major achievements in that journey.

Many developers and companies are using Flynn today in a variety of ways including in production. We’re thrilled to be able to support these deployments technically and organizationally.

The launch of this new stable channel coincides with zero-downtime in-place upgrades of Flynn clusters. The updater itself is in beta and you can follow this guide to update.

We’ll be shipping releases weekly to start, and will adjust the release interval over time.

We also added a page to the docs that explains Flynn’s stability in greater detail.

Recent Changes

Over the past few months we have made several major changes and additions to Flynn that dramatically improve its reliability and performance. Most notably a complete rewrite of the scheduler, switching to a different library to power service discovery (while removing the etcd dependency), and the releases of full cluster backup and cluster first aid tools.

These changes and additions repaired or replaced the components responsible for the vast majority of instability in Flynn clusters. The new tools make it much easier and faster to repair damaged clusters and migrate entire clusters to different infrastructure.

Newsletter

With the release of this new channel a newsletter will now be shipped monthly. Each month we’ll summarize recent developments in Flynn and list upcoming events like meetups and conference talks by the Flynn team and community.

This newsletter will be in addition to our weekly updates on the Flynn blog.

Google Group

There is a new mailing list available on Google Groups to discuss the use and development of Flynn. Issues should still be reported on GitHub and the team will continue to be available in IRC to answer questions synchronously whenever possible. The new list is also intended to be a place for users of Flynn to discuss their questions, experiences and best practices.

Events

Next Tuesday, November 10th we will host a meetup in San Francisco.

Our CTO, Jonathan Rudenberg will demo the most recent version of Flynn as well as some of the newest Flynn utilities like cluster repair and migration. There will also be lots of time for questions and discussion.

We will also be at a number of conferences in the next few months, those dates and details will be announced in future newsletters.

If you would like a Flynn team member to visit your local group for a Flynn talk, email us.

We can also help you put together your own presentation on Flynn if you’d like. Several users have already given successful talks on Flynn at conferences and community events.


As always we want to thank our community and users for their support. We strongly encourage community contributions of code, documentation, and blog posts. We also try to send out T-shirts and stickers to everyone after their first commit or blog post. If you have already made a contribution, but haven’t received a package email us.

We want to hear about any problems, success, pain, or pleasure you’re having with Flynn.

If you have any feedback or questions about Flynn please stop by IRC (#flynn on Freenode), join the Google Group, file issues on GitHub, or email our team any time.

–The Flynn Team

This Week in Flynn

This past week we continued our focus on production stability. A number of bug fixes and enhancements landed to push Flynn further towards that goal.

Flynn’s log aggregation service now operates in a highly available cluster to ensure that log data is not lost.

Flynn’s cluster domain can now be modified after bootstrap via the CLI. Usage: flynn cluster migrate-domain new.domain.com. Remember that DNS entries must be updated accordingly.

Thanks to community member tombh, Flynn now supports HTTP routing for multiple process types of an app.

§ Changes

§ Enhancements

  • Make more Procfile process types routable via HTTP (#2002)
  • HA/recoverable log aggregation (#2008)
  • Migrate cluster domain (#1957)
  • test: Add branch to CI index (#2045)
  • cli: Route export/import (#2025)

§ Bugfixes

  • script: When using multi-host cluster use separate log directories (#2053)
  • test: Handle closed event stream in TestDomainMigration (#2051)
  • installer: Save SSH key used in Azure install (#2050)
  • scheduler: Implement host sync and checking (#2041)
  • scheduler: Retry SyncJobs with a small delay if it fails (#2037)
  • worker: Ensure old jobs are stopped in one-by-one deployments (#2035)
  • cli: Attempt to get cluster details from git before using default (#2033)
  • cli/config: Fix multiple cluster conflict error (#2032)
  • scheduler: Logging improvements (#2029)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

The Flynn team continues to focus on production readiness. This past week saw a number of improvements to stability along with some big enhancements.

The Flynn host daemon now supports zero downtime updates. Soon the updater will be able to update entire clusters without downtime.

Logging has been added or enhanced in a number of components to better track down remaining issues.

We also added an FAQ section to the docs to address the most common questions. If you have a question that isn’t answered there, please open an issue.

§ Changes

§ Enhancements

  • appliance/postgres: Add DATABASE_URL to postgres provider (#2024)
  • Docs: FAQ (#1996, #2023, #1998)
  • router: Replace martini with httprouter (#2009, #2010)
  • host: Support zero-downtime updates (#1672)
  • router, Godeps: Switch to github.com/vanillahsu/go_reuseport (#2001)
  • Discoverd improvements (#2000)
  • controller,cli: Update route (#1980)
  • discoverd: add instance expiration (#1990)
  • cli: Add missing limit command to main help (#1989)

§ Bugfixes

  • host: Add better logging to child exec (#2028)
  • discoverd/server: Don’t log during expiry when not leader (#2022)
  • discoverd/server: Bump raft apply timeout to 30 seconds (#2016)
  • Godeps: Remove go-bindata symlink test data (#2021)
  • dashboard: Add back support for form-urlencoded login request (#2019)
  • discoverd debug logging (#2015)
  • discoverd,scheduler: Clarify nil leader events (#2012)
  • discoverd/health: Panic with error if accept fails in test (#2006)
  • host: Don’t run nsumount without mountpoints (#2007)
  • controller: Misc prepared query cleanups (#1992)
  • discoverd: remove error when no instances need expiration (#1997)
  • Dashboard optimization (#1994)
  • controller: Misc prepared query cleanups (#1992)
  • pkg/httphelper: Retry on net.OpError or syscall.Errno (#1991)
  • slugrunner: Retry downloading slugs (#1987)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

We continued our push towards production readiness this past week as we shipped a number of important bugfixes and enhancements.

Flynn can now be bootstrapped with existing TLS details, e.g. those of an existing cluster, which means a cluster can now be replaced without having to import new certificates on every client.

§ Changes

§ Enhancements

  • dashboard: Replace martini with httprouter (#1985)
  • host: Add cli command for listing hosts (#1975)
  • test/cluster: Add dashboard domain to local DNS (#1971, #1982)
  • bootstrap: Allow passing pre-generated TLS details in env vars (#1970)
  • all: Add Joseph to MAINTAINERS (#1972)

§ Bugfixes

  • bootstrap: Check if ports are available prior to bootstrap (#1986)
  • bootstrap: When checking status in bootstrap, retry if unhealthy (#1983)
  • util/release: Deference device names when updating EC2 manifest (#1979)
  • bootstrap: Allow 30 seconds for jobs to come up (#1978)
  • Fix pinkerton cleanup (#1969)
  • host: Don’t display EndedAt if zero (#1968)
  • release: Restore eu-central-1 (#1967)
  • bootstrap: Report service status details in bootstrap (#1964)
  • Remove goamz (#1951)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

Many bugfixes and enhancements landed this week as part of our effort to stabilize Flynn for production use. Most significantly Flynn now uses pgx instead of pq to talk to PostgreSQL, which improved stability and speed.

§ Changes

§ Enhancements

  • dashboard: Make app history items actionable (#1913)
  • bootstrap: Default min-hosts to len(peer-ips) when appropriate (#1930)
  • host: Update containerinit stream logging (#1928)
  • scheduler: Added issue numbers to TODO comments (#1923)
  • Implement Scheduler V2 (#1914)

§ Bugfixes

  • host: Fixed log fallback command in collect-debug-info (#1925)
  • cli: Fixed panic when force-adding cluster that conflicts with multiple clusters (#1926)
  • slugbuilder: Allow buildpack build to access client SSH keys (#1918)
  • dashboard: Fixed scale app (#1910)
  • schema: Fixed app JSON (#1909)

§ Stay in Touch

§ How You Can Help

 RSS



Mailing List