1.12.0

Features

Support for Keda 2.0

From this release, the message queue trigger of --mqtkind keda, supports Keda 2.0 we won’t be supporting Keda 1.5. Checkout out compatibility page for more details.

Multi-Architecture Build

Thanks to cerebrate, now you can deploy fission on ARM architecture.

Fixes

  • We did a couple of optimizations around choosing the ready pods for specialization by introducing a controller and a queue to avoiding a costly call to k8s API every single time specialization request comes. After the changes, we were able to scale up to 7.5K RPM from 4.8K RPM for long-running functions. In our load test, we used functions having execution time of 15s, 45s, 75s, 90s, and 120s. Link to functions samples. Refer to PR

  • Improved scale performance issue when pool size is smaller than request burst by changing queue implementation strategy. Issue | Fix

  • When using fission spec apply, spec validations are now made optional which performance by twice. Issue link | Fix

  • Earlier when the number of requests exceeded the value of the concurrency parameter, 502 HTTP status code was propagated to the user. Now, instead 499 HTTP status code would be propagated. Issue | Fix

  • When a function(new deploy strategy) and environment were deployed in a different namespace, changes made environments weren’t reflected in the function pod. Now, this is fixed. Issue | Fix

  • While working with fission spec apply --wait build status of package wasn’t checked, this has been fixed now and exit code 1 is returned on the failure of a package build. Issue | Fix

  • Replaced clusterRole cluster-admin with a less permissive role. Issue | Fix

  • Thanks to chenjr0719, now error message is put to error topic only after max retries in Kafka MQT of --mqtkind fission. Issue | Fix

Last modified February 15, 2021: Release 1.12.0 doc update (#242) (2514362)