OSprofiler is a lightweighted OpenStack-centric profiling library, that provides functionality to generate one trace per API request, that goes through all involved services and clients. This trace can then be used to build a tree of calls which can be quite handy for a variety of reasons (e.g. in isolating cross-project performance issues).
During a long trip from PoC to Oslo library, OSprofiler has been interated to lots of OpenStack projects (Nova, Neutron, Cinder, Glance, Heat, etc.), allowing to evaluate OpenStack behavior and find issues on REST API, RPC and DB usage layers. Newest driver model allows to store API requests profiling information in various storage options - starting with classic OpenStack-telemetry to MongoDB, Elasticsearch, Redis and Log Insight solutions.
Please join us on this presentation to learn more about the library, its architecture and possible use cases, that can make OpenStack development and debugging step on the new level.
- Reasons to create a specialized library for OpenStack profiling
- Use cases, tips and "how to"s
- OSprofiler architecture: what is profiled and how
- Adding OSprofiler support to the new OpenStack ecosystem project
