Current Series Release Notes¶
- Allow to search for values in metrics by using one or more granularities.
- The REST API authentication mechanism is now pluggable. You can write your own plugin to specify how segregation and policy should be enforced.
- Add support to backfill timestamps with missing points in a subset of timeseries when computing aggregation across multiple metrics. User can specify fill value with either a float or null value. A granularity must be specified in addition to fill.
- When sending measures in batch for resources, it is now possible to pass create_metric=true to the query parameters so missing metrics are created. This only works if an archive policy rule matching those named metrics matches.
- Add resample parameter to support resampling stored time-series to another granularity not necessarily in existing archive policy. If both resampling and reaggregation parameters are specified, resampling will occur prior to reaggregation.
- The gnocchi-config-generator program can now generates a default configuration file, usable as a template for custom tweaking.
- A healthcheck endpoint is provided by default at /healthcheck. It leverages oslo_middleware healthcheck middleware. It allows to retrieve information about the health of the API service.
- gnocchi-metricd now uses the cotyledon/oslo.config helper to handle configuration file reloading. You can dynamically change the number of workers by changing the configuration file and sending SIGHUP to the metricd master process.
- New storage driver for AWS S3. This new driver works in the same way that the Swift driver, expect that it leverages the Amazon Web Services S3 object storage API.
- The new auth_type option specifies which authentication system to use for the REST API. Its default is still noauth.
- Ceph driver has moved the storage of measures metadata from xattr to omap API. Already created measures are migrated during gnocchi-upgrade run.
- The api.pecan_debug has been removed.
- The created_by_user_id and created_by_project_id field are now deprecated and being merged into a unique creator field. The old fields are still returned and managed by the API for now.
- ceph driver now uses the rados async api to retrieve measurements to process in parallel.
- The default archive policies “low” and “medium” are now storing less data than they used to be. They are only using respectively 1 and 2 definition of archiving policy, which speeds up by 66% and 33% their computing speed.
- Gnocchi now leverages microseconds timestamps available since MySQL 5.6.4, meaning it is now the minimum required version of MySQL.
- The default archive policies list does not contain the 95pct and median aggregation methods by default. These are the least used methods and should make gnocchi-metricd faster by more than 25% in the default scenario.
- A new REST API endpoint have been added to be able to update a resource-type: “PATCH /v1/resource-type/foobar”. The expected payload is in RFC6902 format. Some examples can be found in the documentation.
- The Carbonara based storage engine has been updated and greatly improved. It now features fast write for Ceph (no change for file and Swift based drivers) by using an append method. It also features on the fly data compression (using LZ4) of the aggregated time serie, reducing the data space usage by at least 50 %.
- Swift now supports authentication with Keystone v3 API.
- gnocchi-upgrade must be run before running the new version of gnocchi-metricd and the HTTP REST API in order to upgrade from version 2 of the Carbonara storage engine to version 3. It will read all metrics and convert them to new version 3 serialization format (compressing the data), which might take some time.