Cloud Foundry Advisory Board Meeting, Feb 2016: Summit CFP Is Now Open!
The Cloud Foundry Summit 2016
The Cloud Foundry Summit 2016 Call for Papers (CFP) is now open, and will remain so until February 29, according to an announcement at the monthly Cloud Foundry Community Advisory Board (CAB) conference call.
The Summit will be held at the Santa Clara Convention Center and adjoining Hyatt Regency in Santa Clara, CA. Organizers from the Cloud Foundry Foundation (CFF) expect around 2,000 attendees this year.
The CFF’s is urging potential speakers to focus on submitting a single submission—“your best submission”—noting that the CFF review team can become inundated with submissions. He expects the final program to be a mix of hands-on technical sessions along with interesting use cases and customer stories.from IBM moderated the call as usual, with perhaps 20-odd participants (compare and contrast with 20 odd participants.)
Several technical reports came from project team managers. Please note that several links to detailed information about a lot of the bullet points below can be found in the CAB call agenda at lists.cloudfoundry.org. People who are interested should subscribe to the lists if they haven’t already.
MEGA ( , Pivotal)
Amit reported on infrastructure and release integration.
Infrastructure (Consul, etcd, NATS, Postgres)
- Working on Consul 1.5.3 implementation. There are many improvements to robustness of consul and etcd releases.
- Much refactoring of Consul to remove more logic out of bash.
- Much refactoring of consul-acceptance-tests and etcd-acceptance-tests to isolate tests from external dependencies.
- Onboarding two new team members.
Release integration (cf-release, integration pipelines)
- Four new releases cut.
- Separated out several credentials repos to mitigate risks associated with leaks.
- Continued work with Diego to merge DATS into CATS and get a single suite of tests against Diego and DEA.
- Significant progress on route-registrar agent component, supporting sophisticated health-check configuration, preparing handoff to routing team.
- Cut up to release v230.
- Lots of issues, questions, and Prs, and lots of pipeline maintenance, especially with shepherding through CVE fixes.
- Helped with Java buildpack fix.
- Wrapping work on route-registrar component to create common route-reg behavior.
Loggregator ( , Pivotal)
- Finishing up Doppler -> Metron TCP (to go with previously released TLS).
- Working on Metron throughput.
- Incepting on Executor -> Metron TCP/TLS Thursday.
CLI ( , Fujitsu)
Dieu Cao of Pivotal made the report on behalf of Dies Koper, who is based in Australia and was not on the call.
- Close to releasing CLI v6.16.0, which will expose routing services, space-scoped service brokers, and a number of community requested features and bug fixes. Originally this release was to expose TCP routing features, as well, but decided to cut a release now already as there is still a lot of work ahead for TCP routing.
- The first v3 API story (cf v3apps) was delivered. Not all required backend functionality was implemented, so, created a new story to complete this command blocked on CC. Also shelving this epic until we complete the routing epics and CAPI is ready to focus on the API features our commands need.
- Completed (minor) revamp of CLI chapters on docs.cloudfoundry.org. Removed CLI v5 references, also rearranged order of sections, so similar topics are better grouped.
- Restarted GoCD to Concourse migration as we’ve been experiencing instability with our linux32 CI box.
- Working with the Doc Tools team to integrate a command reference guide generated from CLI help output: basically, cf help output will be integrated into CLI docs above and each command will hyperlink to an ‘external’ site (like apidocs) with pages with generated help from all commands and supported languages.
- Looking into strategies to keep the language files up to date (no progress).
- Working on a proposal for a cli command to check if it’s the latest, potentially update itself (no progress).
Routing ( , Pivotal)
Shannon reported on a TCP routing support milesteone using cloud controller API to map TCP route to an app and specfic app port. Said there are still some gaps to fill in use cases.
The team working in parallel with the CLI team to implement support for multiple app ports and TCP routing. He confirmed that the CLI team intends team to ship 6.16 with support for route services.
- You can now map routes to apps and specify an app port.
- You can specify routes to apps on multiple ports.
- TCP routing is now usable (supported on Diego only). Requires cf-routing release to be deployed alongside cf-release.
- Working with the CLI team on adding support for these features.
- xtp with Diego to mitigate the possibility of stale routes.
BOSH ( , Pivotal)
- Polishing up feature work on links v2, adjusting based on feedback.
- Producing stemcells for CVEs.
- Trying to remove Redis and upgrade Postgres to a newer version.
- Finished implementing addons(configurations for the director to allow specifying a job to be collocated globally), will merge from the same branch as links.
- Continue to migrate from Jenkins to Concourse, stemcell building and tests to Concourse -> publishing sha1s for stemcells.
- CPI team finishing up vSphere features to support VMotion.
- Changing some of the pipelines to certify stemcell versions against BOSH release and CPI versions.
- Released bare metal CPI—rack hd with EMC.
- Minor changes to Azure CPI, based on feedback to take advantage of Azure features.
CAPI ( , Pivotal)
- Continuing to make progress on support for Tasks. Currently fixing up auditing, quotas, and orphan mitigation.
- Continuing work on v3 and application process types—adding support.
- Very experimental new API docs for V3 now posted online.
- Work on replacing the nfs included in cf-release with webdav for the default blobstore is nearly complete and should be available to opt-in in a soon to be available cf-release and once we’ve had a bit more burn-in in test environments we’ll be switching it to default.
- Instructions for required manifest changes posted.
- Will no longer support nfs included in cf-release, but will continue to support externally provided nfs.
- The blobstore release is still a single point of failure and we recommend not using it for environments requiring HA of all components.
Later, she also mentioned that the team is introducing app usage events separating out staging as separate billing app usage events. Dr. Max voiced his approval and noted this will change the way product teams are thinking about how they’ll charge.
Dieu also mentioned that she is lobbying the Tracker team for better integration between Tracker and Github, and is focused on prioritizing to allow viewers in tracker projects to follow stories.
Diego ( , Pivotal)
- Added rep bulk loops to BBS benchmark tests.
- 200K records across 1000 reps too much for naive BBS read policy against etcd.
- Investigating potential solutions: BBS in-memory record cache; relational store for BBS DB—a lot of our data i pretty relational—can make change transparently.
- Auctioneer changed to distribute in-flight container starts more evenly.
- Suppressed ‘noisy’ health-check log output, logging only health transitions.
- Finished merging DATs into CATs, switched Diego CI to run only CATs, and will remove DATs after CF v231 released.
- Documenting and publicizing team tooling to deploy BOSH+CF+Diego to AWS.
- Now have someone in Denver to work with Loggregator team to enable executor-to-metron TCP communication.
- Will install BOSH-delivered trusted certs in cflinuxfs2 rootfs, provide to all Linux containers.
- Various improvements to logging, metrics, docs, and manifest generation.
Persi ( , Pivotal)
Ted is the project manager for the new persistence team, dubbed “persi.”
- Picked CephFS as our reference implementation. Stable and easy to re-deploy so doesn’t gunk up CI.
- Reviewing Flocker and REX-Ray from ecosystems (Docker and Mesos implementations) for architecture choices, potential for sharing driver ecosystems and core libraries.
He also mentioned an interest creating a shared standard, rather than adopting another product as an implementation, and getting “a little pressure” from the Docker community.
Dr. Max asked whether it makes sense to involve the Linux Foundation with respect to creating a standard for persistence with PaaS or Cloud. Chip said it absolutely made sense and noted that the CFF is a part of the Cloud Native Computing Foundation. He said the CF community should get involved “and see if we can make a useful standard.”
Ted then added that the Persi team would love to learn more about what sorts of distributed file systems the CFF members are using and how they’re using them, so that the team can serve members in this area.
Garden (Dr. Julz, IBM)
- Shipped v0.331.0 and v0.332.0 containing various small bug fixes.
- Working on more configurable graph cleanup behavior for v0.333.0. Operators will be able to set a disk usage threshold before which graph cleanup will not be triggered.
Guardian (runC backend)
- Passing all Diego smoke tests apart from Docker smoke tests.
- Hope to have full suite of CF acceptance tests soon and then think about rolling Guardian to production.
- Now have a Diego/CF deployment with Guardian with DATS passing for buildpack based apps. Expect to get rest of DATS passing this week.
- Will soon begin conversations with Diego team about furter integration and testing.
Greenhouse/.NET ( , Pivotal)
- Finally finished with the “tar pit” and swapping out our tar implementation for BSDtar, including switching the app lifecycle./li>
- Completed the Windows side of the Diego bind-mounts work.
- A number of chores and bug fixes, including validating and accepting some pull requests.
- Added a number of checks and tests to Hakim, our “make sure everything is in a good state”-tool.
- Continued dealing with fallout of tar issues (i.e., long filename issue).
- Moving forward continuing with bug fixes, chores, and seeking feedback from real world users.
Buildpacks / Stacks ( , Pivotal)
- Updated buildpacks: Ruby, Go, PHP, Staticfile, Node.js.
- Using Concourse to auto-test pull requests—automatically test it and pass / fail it based on state of the build. A feature we added to something that we built and found some good use for it.
- PHP7 support in progress and sunsetting HHVM.
- “BOSHifying all the buildpacks” story under way to support BOSH release of each one.
- #buildpack-firehose in Cloud Foundry slack now states when builds fail.
UAA ( , Pivotal)
- UAA CVE-2016-0732, privilege escalation
- UAA 2.7.4 & UAA 3.0.1 released
- UAA release 3.1.0 released with whitelabel support, all Pivotal resources and references removed.
- Fixed the issue with Passcode URL (because of login.url removal from the spec file).
- UAA 3.2.0 release work in progress.
- Monit script fixes to address UAA downtime.
- SAML single logout support.
Abacus ( , IBM)
- Reported a very successful presentation at Pivotal.
- Created a series of three blog posts to introduce CF-Abacus. Part 1 is published, the other parts are scheduled. Contact Dr. Max directly for an earlier viewing of parts 2 and 3.
- Started work on the failed event management. Service brokers can cause failed events in Bluemix.
- V3 API is introducing new app usage events to break out staging as a distinct event for use in Abacus.
- Adding process by which events in error state can be modified in bulk.
- SAP added support for MongoDB. Required some refactoring to the data layer. Now testing
SAP also started work on creating Abacus Concourse.
Flintstone / Bits-Service ( , IBM)
- Founded a joint team with two IBM and two Pivotal engineers for the bits-service incubation project.
- Incepted on the bits-service and did a week F2F.
- Created bits-service and bits-service BOSH release project in git (under cf-incubation).
- Established a Concourse pipeline for bits-service into the Softlayer environment.
- Finished extraction of the create-buildpack path from the cloud_controller into the bits service project, usage in a private cc branch controlled by a feature flag.
- Focusing on buildpack initially.
- Deleting and updating events.
- Will test using a separate branch and feature flags in upcoming releases
The next monthly call is currently scheduled for Wednesday, March 9 at 8 a.m. Pacific Time. Potential participants are encouraged to keep track of postings at cf-dev in case this changes.
Also, the calls have migrated online to a Slack channel. People interested in being on the call can register anytime here.