Sharding a time-series collection is enabled.

• The DDL operations for indexes and collMod are supported. The updating a granularity value is currently not supported on sharded clusters. We are planning to support this in 5.3.
• The sharding admins commands can only be run on the buckets namespace.

The shardCollection command accepts an optional timeseries parameter.

• If a timeseries collection already exists,
• Either the timeseries spec of the shardCollection command must match the collection spec.
• Or the timeseries spec should not be specified. In this case the timeseries spec from the collection catalog is inherited.
• If there is no existing collection and the timeseries parameter is specified, a new time series collection will be created.

The project is backported to 5.0.4 with the feature flag off. The feature will be enabled on 5.0.6

TimeseriesOptions:
description: "The options that define a time-series collection."
strict: true
fields:
timeField:
description: "The name of the top-level field to be used for time. Inserted
documents must have this field, and the field must be of the BSON UTC
datetime type (0x9)"
type: string
metaField:
description: "The name of the top-level field describing the series. This field is
used to group related data and may be of any BSON type. This may not
be \"_id\" or the same as 'timeField'."
type: string
optional: true
granularity:
description: "Describes the expected interval between subsequent measurements"
type: BucketGranularity
default: Seconds
bucketMaxSpanSeconds:
description: "The maximum range of time values for a bucket, in seconds"
type: safeInt
optional: true
validator:

Summary

Support sharded time-series collections

Motivation

Add Sharding support for time series collection to allow for horizontal scaling of time series collections. 

Cast of Characters

• Product Owner: Michael Gargiulo
• Project Lead: Arun Banala
• Program Manager: ana.meza
• Drivers Contact:

Documentation

Scope Document
Technical Design Document