The latest Apache Aurora release, 0.22.0, is now available for download. Here are some highlights in this release:
update_strategy. Update strategy may take a
BatchUpdateStrategy, or a
BatchUpdateStrategytake a single integer argument while
VariableBatchUpdateStrategytakes a list of positive integers as an argument.
ROLLING_FORWARD, the update will automatically pause itself right before a new batch is started. (This feature is being released as tested but in beta state. We are looking to collect feedback before we consider it fully stable.)
loader.load()now uses memoization on the config file path so that we only load and process each config file once.
--disable_task_resource_collectionto disable the collection of CPU, memory, and disk metrics for observed tasks. This is useful in setups where metrics cannot be gathered reliable (e.g. when using PID namespaces) or when it is expensive due to hundreds of active tasks per host.
-sla_aware_kill_non_prodwhich allows operators to enable SLA aware killing for non-production jobs. Jobs are considered non-production when they are preemptable and/or revocable.
Deprecations and removals:
JobUpdateSettings.updateGroupSize. Please set the proper
JobUpdateSettings.updateStrategyinstead. Note that these same constructs, as represented in the Aurora DSL, are still valid as they will be converted to the new field automatically by the client for backwards compatibility.
Full release notes are available in the release CHANGELOG.
Thanks to the 8 contributors who made Apache Aurora 0.22.0 possible: