Defining Partitions
To have partitioned statistics, your application must have resources.
Partitions are a stat-mapping
schema, and the SSID has a stat-mapping
component instead of a stat
component. Partitions are named in a partitions
element within an aggregations
or instance-metadata
element in the metadata file, as shown in the following example:
"$schema": "//:stat-mapping", "id": "//:class.app/util2//:stat-mapping.errors", "instance-metadata": { "partitions": [ "inst" ] }, "transforms": [ { "match": "//:class.app/util2//:res.inst/(inst[1-2])//:stat.errors$", "replace": "//:class.app/util2//:stat.errors//:part.inst(\\1)" } ]
In the transforms
section, each match
represents a statistic that is already defined in the class
and stat
metadata, and the paired replace
maps that statistic to one of the partitions named in the partitions
element. The value of the partitioned statistic is equal to the sum of the values of the matching statistics for each resource.
See Separating Data Into Partitions for an example. See also the "Statistic Mapping" section in the ssid-metadata
(7) man page.