cloudfoundry/diego-release v1.31.0

Diego v1.31.0

on GitHub

Changes from v1.30.1 to v1.31.0

• Verified with garden-runc-release v1.10.0.
• Verified with garden-windows-bosh-release v0.12.0.
• Verified with etcd-release v117.
• Verified with cf-mysql-release v35.
• Verified with cflinuxfs2-release v1.170.0.

IMPORTANT: Updated 2017-12-13. We have identified an issue with the calculation of the memory container metrics in this version of Diego for those cells that do not enable the container proxy (which is the default configuration). For apps with a memory limit under 18 MB, the miscalculation results in reporting a very large number for the container memory usage, which causes cf push to fail when checking on the app instance status with an error of the form Could not fetch instance count: Invalid JSON response from server: json: cannot unmarshal number 18446744073708089344 into Go struct field .Mem of type int64. We are currently working to address this issue in story #153655210 and will include the fix in a new final Diego release as soon as possible.

As a workaround, operators should set the containers.proxy.additional_memory_allocation_mb property to 0 on the rep and rep_windows jobs on their Diego cells. We will also submit changes to cf-deployment to add this workaround automatically. We apologize for the inconvenience.

Significant changes

BBS API

• When an app crashes and the BBS auction request fails, the rep crash-actual-lrp should not fail

Container Placement

• As a CF operator, I expect to be able to override the BOSH-provided availability-zone value so that I can define "logical" availability zones that differ from the IaaS-provided ones

Routing

• route-emitter should be able to handle actual lrp update events during sync

Declarative Health Checks (Experimental)

• As a CF operator, I expect the long-running healthcheck processes to run in sidecar containers with large enough no memory limits so that they do not cause memory-wedged containers on Linux

Per-Instance Proxy (Experimental)

• As a Diego operator, I expect the default additional proxy memory allocation to be updated to account for the LDS-server overhead so that I can account for memory allocations accurately
• As a CF operator, I expect enabling container proxies on Linux cells to supply all LRP containers with proxies so that I can provide route integrity to all traffic ingressing to containers

Component Logging and Metrics

• Operator should be able to see how many route registration/unregistration messages are emitted by NATs for internal routes in their environment
• cloudfoundry/route-emitter #11: Add new metric for number of internal routes emitted

Dependencies

• cloudfoundry/clock #4: Don't allow fake tickers with non-positive periods.

Documentation

• As a Diego contributor, I expect scripts/check-metrics-documentation to work with the new loggregator client so that I can be sure component metrics are documented adequately (in flight)

Licensing

• bump https://github.com/jeffpak/local-node-plugin/ to include Apache license

BOSH job changes

None.

BOSH property changes

rep and rep_windows

• Default value of containers.proxy.additional_memory_allocation_mb changed from 5 to 18.

BOSH link changes

None.