Cloud Foundry Advisory Board Meeting, January 2016

Though the holidays had some anticipated effect on various aspects of progress, the meeting nonetheless reflected many positive steps over the past month.

Dr. Max (Michael Maximilien) from IBM resumed his normal role as moderator, after having been in China and unavailable in December. He took attendees through a brisk survey of recent progress.

 
Cloud Foundry Foundation Updates ( , )

• Chip thanked all attendees for an outstanding 2015. “It’s been an amazing ride so far,” he noted, and is looking forward to more progress in 2016. He cited Industrial Internet projects from GE Software, Bosch, and Siemens as particular highlights, along with national government projects in the US, UK, Australia, and South Korea.
• Chip also said that the call for papers (CFP) for this year’s Cloud Foundry Summit (to be held in Santa Clara, CA May 24-25) will be open before the next CAB call in February.

• Stormy announced two mailing lists: cf-speakers and cf-meetup-organizers. She also noted the availability of “care packages” for meetups containing t-shirts with the new Cloud Foundry logo, and reiterated the presence of the cf-speakers list for people who are interested in speaking at various meetups throughout the world.

MEGA ( , Pivotal)

• The infrastructure team fixed some regressions in Consul.
• Rewrote startup scripts from Bash to Go.
• Now building fault-tolerance tests for Consul and running them in the CI with FCD.
• Investigated bug reports causing the Diego DB VM to fail to come up.
• On the release integration side doing a lot of work on cats, mostly a bunch of cleanup to make sure they are tearing down resources they create, and integrating few other best practices.
• In the common route registrar, adding features that the UAA team had requested concerning deregistration and health checking. Rewrote from Ruby to Go and writing some tests.

UAA ( , Pivotal)

• Some major refactoring for 3.0 release, bringing server and client sides to separate packages.
• Issue of white labelling for UAA—looking for proposal on mailing list.

GitHub repo

Loggregator ( , Pivotal)

• Had some positive functionality release.
• Metron->Doppler TCP/TLS link released.
• Loggregator as a separate BOSH release (consumable by non-CF BOSH deployments) released. There’s a readme on the Loggregator github repo with details.
• Loggregator performance characterization is ongoing.
• There’s an ongoing effort to provide app-to-nozzle reliable chains. This won’t be completed by the next CAB meeting, but work will be underway.

GitHub repo

CLI ( , Fujitsu)

• Dies was not on the call but provided a progress list for the month.
• Released v6.14.1 in Dec and v6.15.0 this week, with support for routing with paths and IBM contributed translations for CLI help and messages for nine(!) languages.
• Continuing to work closely with Shannon & Routing team on exposing Route Services and TCP Routing features through the CLI.
• Upcoming work is concerned with exposing more Routing features, such as configuring multiple app ports.
• A Pivotal engineer from San Francisco is visiting Sydney (Australia) this month to ramp up a developer from Fujitsu who joined the CLI team.
• Also having a Sydney-based Pivot remote cross-pairing with the Routing team on stories to expose Routing features through the CLI.

GitHub repo

Routing ( , Pivotal)

• Support for route services on backend nearly done, with one story left.
• API should be complete. The CLI team has been engaged and helpful in prioritizing routing stories.
• Will be releasing veson of CLI this week with support for routes with path
• Wrapped end back-end support for multiple ports for an app.
• Working on most important epic, which is mapping a TCP route to an app and specifcying an app.
• Also had some improvement regarding the use of tokens.
• Have been instrumenting components “a lot.”

Services Core ( , Pivotal)

• MySQL release v25 to cf-dev mailing list. Information can be found there.
• The big thing is the ability to do much less manual bootstrap process. Run a BOSH errand “and it works 80% of the time.”
• Also getting community feedback from a posted survey.
• Close to releasing v26, with an important coding enforcer fix.
• The other big feature is the ability to save 30% of platform spend. This has to do with fully replicated clusters, which need three nodes. When MySQLscales vertically costs can run into the thousands of dollars. Now there’s a feature called arbitrator, which allows data replication to two big nodes and one tiny node.

BOSH ( , Pivotal)

• There’s been a lot of work delivered but not quite announced.
• The team’s trying to get to the point where requirements are “more or less automatically deduced” for particular jobs.
• Working on orphan disk scenarios to avoid deletion of disks, as a production safety feature. The side effect is to speed up deployment.
• Working on “add-on” feature called runtime config.
• Working on links v2. Coming in a week or so.
• Improving CPI behavior.
• Working on some AWS API improvements, including how to build stemcells in AWS.
• Started putting stemcells for Azure into BOSH I/O.

GitHub repo

CAPI ( , Pivotal)

• Accepted Persistence team into incubation.
• Runtime PMC proposal is on cf-dev and accepted Networking team.
• Continiung work with CAPI on process types. The Java client library is providing feedback on many endpoints.
• Making good progress on replacing NFS with web-dot cloud protocol. Users won’t need to do a blog migration, but instead just use an existing VM and put their web server there.
• With cf-release 227 there are additional stories needed to be done.
• CLI support to be coordinated with Dies and team.
• Haven’t removed experimental side yet but will do.
• Removed support for v1 service broker.

GitHub repo

Diego ( , Pivotal)

• Delivered a couple of 1446 releases.
• Bind map downloads including buildpacks.
• Finished upgrade test suite, and have already caught a couuple of things.
• Re-enabled BBS benchmark tests and dramatically imrpoved their performance.
• Have been investigaing some slow evacuation times that were observed at higher density deployments. This is related to how much disk space is available on VMS.
• SSH now suports ttcp. Addressing reports of logging around health checking being noisy, so muting individual health checks.
• Taking Diego acceptance tasks and merging into a unified suite.

GitHub repo

Garden / Guardian ( , Pivotal)

Garden-Linux

• Fixed some bugs relating to use of custom Docker registries as the default registry
• Turned on noatime and turned off journaling for the graph filesystem in an attempt to improve staging performance.
• Have not yet measured the impact due to CI pipeline issues.
• Will ship v0.331.0 with the above changes once CI is back to normal.

GitHub repo
 
Guardian (OCS/runC)

• Now running the Diego smoke tests against a CF deployment using Guardian.
• Next step is to use Vizzini (Diego tests) and DATS + CATS to confirm prioritization of remaining work required to migrate Diego over to Guardian. Working towards getting a version ready to ship.

Pivotal tracker – GitHub repo

Greenhouse/.NET ( , Pivotal)

• Fixed some bugs, and working with Diego team to get cross-team pipelines set up.
• “Determining when one of us breaks the other.”
• Cleanup chores and one-off features to get back up to speed.

GitHub repo

Buildpacks/Stacks ( and , Pivotal)

• Provided Full PMC Notes from 2015-10-12.

• Danny is taking over day-to-day management of buildpacks. It was noted last month that he was in CF Dojo training and would be expected to assume this role early in 2016.
• Supporting wildcards in patch releases.
• Node.js support supporting table, continuing to upgrade buildpack.
• Added 344 support ot python updates as well for PHP and Ruby.
• Reviewing an IBM pull request for Power PC support.

Abacus ( , IBM)

• Usage collection
• Handle runtime usage GUID resets.
• Metering and rating configuration at the resource/plan level.
• Usage rating at the resource instance level.
• Remove separate rating service.

• Usage accumulation and aggregation
• Include account id in aggregated usage.
• Remove deprecated region property.
• Optionally post aggregated usage to an external service.

• Deployment
• Make it a little easier to deploy.
• Upgrade to latest Node dependencies.
• Improve BOSH release creation and deployment scripts.

• Bug fixes and improvements
• Fixes to time window and slack window processing logic.
• Improve memory usage and fix memory leaks.
• Tune throttling and cache sizes.
• Disable BigNumber errors with more than 15 significant digits.
• Don’t write passwords to debug log.

Persistence ( , Pivotal)

• Adding a volume manager interface in Diego cells. Idea is to mount external devices inside containers.
  Plan to support multiple device managers simultaneously.
• Support all kinds of persistence to use block device mounted into container.
• It’s a runtime component. Will try to do so in an industry-standard way as standards emerge in the world of persistence.
• Working with CAFI and services to expose some form of persistence to existing ephemeral app model.
  Looking to allow services such as Watson to provide device in a container.
• Working to have services that could provide these things and binding services to apps—getting persistence into third-party apps like WordPress, LAMP stack to run as regular ephemeral apps.
• There’s a proposal to spin up a CF team to implement the Persistence proposal

Health Manager ( , IBM)

• Upgrading runtime og (Note: There was speculation on whether “og” means “original gangster” or something else. This writer thinks it may mean “older generation.”)
• Health manager 9000 moving to latest version of Go
• Moving off collector for consistent way of saving their metrics

As always, the project managers noted they are available and open for discussion on issues, and can also often refer inquiries to ongoing reports and postings on github and cf lists. The next CAB meeting is scheduled for Wednesday, February 10 at 8 am pacific time.