RSS

This Week in Flynn

This week the Flynn team continued work on stability and performance.

One big feature landed: Path-based routing is in the latest release, which enables multiple apps to use the same domain. To use it, simply specify a route with a full path in it, e.g.: flynn -a test route add http example.com/test/, or to use a different service in the same app: flynn route add http --service assets example.com/assets/.

§ Changes

§ Enhancements

  • controller: Remove “crashed” and “failed” job states (#2265)
  • Path based routing (#2233)

§ Bugfixes

  • scheduler: Rectify omni job counts when a host goes down (#2282)
  • host/cli: Disable log round-trip during flynn-host run (#2284)
  • controller: Don’t return null meta when creating app (#2285)
  • controller: Update app endpoint now handles setting empty meta (#2267)
  • cli: Remove legacy SSH git code (#2280)
  • schema: Fix event schema (#2283)
  • scheduler: Don’t log triggers (#2281)
  • controller: Ignore deleted apps when inserting new resource records (#2279)
  • Update API docs (#2278)
  • installer: Prompt for credentials on auth failure (#2251)
  • host: Add jobs to state as early as possible (#2270)
  • pkg/syslog/rfc6587: Fix parsing of lines that are the maximum length (#2273)
  • logaggregator: Add missing GET /snapshot endpoint (#2274)
  • host: Retry creating AUFS mounts (#2272)
  • test: Don’t deploy discoverd in single node cluster (#2271)
  • logaggregator/buffer: Skip duplicate log lines (#2269)
  • controller: Fix validation error return when getting active formations (#2264)
  • bootstrap: Populate flynn-system-app meta on jobs before controller start (#2244)
  • installer: Close modal after adding credential (#2253)
  • installer: Add redirect for /clusters -> / (#2252)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This past week we continued to work on stability and performance.

The cluster backup now has an API endpoint, and bugfixes landed for the cluster updater.

§ Changes

§ Enhancements

  • installer: Move prompts out of modal (#2239)
  • controller: Add cluster backup endpoint (#2115)
  • Make updater aware of versions (#2207)

§ Bugfixes

  • test: Give initial VM more memory for running tests (#2247)
  • controller: Copy app metadata to host jobs (#2243)
  • all: Fix CORS handler usage (#2231)
  • util/release: Strip leading “v” from version in vagrant manifest (#2234)
  • Fix TUF segfault (#2216)
  • discoverd: Close mux before closing store (#2212)
  • installer: Remove max number of instances (#2210)
  • controller: Fix hang in controller tests (#2208)
  • docs: Fixed typo (#2206)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

This week we continued to fix bugs and focus on stability and performance.

Several bugs were fixed in the scheduler. Discoverd also no longer persists heartbeats to disk which dramatically reduces the writeload on nodes.

§ Changes

§ Enhancements

  • Update discoverd (#2003, #2183)
  • pkg/status: Return version in status (#2199)
  • discoverd: in-memory heartbeating (#2194)
  • flannel: Pull source code into repo (#2197)
  • Scheduler Updates (#2171)
  • discoverd: mux (#2174)
  • test: Trace commands in runner Upstart script (#2184)
  • host: Add configurable tags, broadcast via service discovery (#2154)

§ Bugfixes

  • controller: avoid panic when streaming formation updates (#2177)
  • controller: Don’t call c.Assert in goroutine (#2202)
  • test: Run domain migration tests after other tests (#2201)
  • scheduler: Fix data race accessing s.jobs (#2198)
  • installer: Don’t show size options with < 1GB ram (Digital Ocean) (#2192)
  • pkg/knownhosts: Fix parsing file with comments (#2191)
  • bootstrap: Correct discoverd port in manifest template (#2187)
  • router: fix test build failure (#2186)
  • test: Stream logs before signalling partial logger (#2181)
  • deployer: skip rollback on all-at-once if old jobs failed to stop in timely fashion (#2180)
  • host: remove bootstrap double prints (#2170)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

A number of bug fixes landed this past week, including a fix for an intermittent time out in Flynn’s installer.

§ Changes

§ Enhancements

  • Add raft log cache (#2145)
  • cli: add app info command (#2107)

§ Bugfixes

  • gitreceive/receiver: Set service on process type for zero-downtime deploy (#2165)
  • controller: Convert from enums to tables (#2153)
  • controller: Always connect job event stream during deployment (#2160)
  • script: Update usage in export-components (#2163)
  • test: include custom port on ci links (#2146)
  • controller: Fix dashboard URL env when migrating domain (#2149)
  • installer: Increase number of ssh connection attempts for bootstrap (#2152)
  • logaggregator: do not send backlog when requested lines are zero (#2123)
  • discoverd: client keep headers on redirect (#2142)
  • pkg/httphelper: Fix retrying on pgx.PgError (#2141)
  • Godeps: Bump github.com/pkg/browser (#2140)
  • test/runner: Increase timeout for integration tests to 25m (#2137)
  • controller: Add App.DeployTimeout (#2135)

§ Stay in Touch

§ How You Can Help

This Week in Flynn

Since our last update the team announced the Flynn stable release channel and we shipped a number of bug fixes and new features.

Flynn clusters can now be fully backed up and restored via flynn cluster backup and flynn-host bootstrap --from-backup respectively.

There is also a new cluster repair tool which can fix several common cluster problems. Run flynn-host fix from a Flynn host to start it.

§ Changes

§ Enhancements

  • bootstrap: Set PG* env vars on postgres release (#2112)
  • status: Require key for status endpoint from public IP (#2103)
  • Update docs (#2079, #2080, #2081)
  • Full cluster backup/restore (#2055, #2088)
  • Support multiple release channels (#2064)
  • script: Support installing explicit version (#2105)
  • docs: Remove redundant nav sub-headings (#2081)
  • Cluster repair tool (#2072)

§ Bugfixes

  • host,test: Fix log collection fallback (#2131)
  • logaggregator: Use retry dialer in client (#2132)
  • Makefile: Fix git tag determination (#2130)
  • deployer: Wait for job events before deleting formation during rollback (#2049)
  • dashboard: Deploy bug fixes (#2128)
  • script: Upload latest snapshot in release-timestamp (#2119)
  • dashboard: Include backlog when viewing app logs (#2113)
  • worker: Always scale old formations down after deployment (#2077)
  • installer: Use a separate EBS volume for data (#2063)
  • script: Check commit has passed CI before tagging (#2062)
  • util/packer: Wait 60s before running provisioning scripts on EC2 (#2040)
  • script: Fix multi-host router IP to match new IPs (#2090)
  • discoverd: Fix test log message (#2104)
  • cli: support flynn cluster add when git is not present (#2094)
  • controller: Inject Flynn app name to the apps ENV vars (#2093)
  • appliance/postgresql: Fix test discoverd instance metadata (#2083)

§ Stay in Touch

§ How You Can Help

 RSS



Mailing List