Query processing code

[Monika Shah]

How can we trace query processing as well as query optimization?

[David Storch]

Hi Monika,

I'm not sure exactly what you mean by tracing query processing, but you could try enabling verbose logging for the QUERY log component. This will log lots of detail from the query planner in particular. You can enable this by connecting to the server with the mongo shell and running the following:

db.setLogLevel(5, "query");

If you provide some more detail about what you were trying to do, I might be able to be more helpful!

Best,

Dave

[Monika Shah]

Thank you David.
I wanted to ask you that how can I trace source code for query processing.
I want to know how can I contribute for performance metrics. So, I want to understand path and apply change for testing my idea

[David Storch]

Hi Monika,

There are quite a few submodules involved in query processing: the query planner, the query execution engine (composed of the query exec library and the aggregation framework libraries), the full text search engine, geospatial search, the mongos query execution path, and so on. In order for me to help you find a starting point for reading the source code, could you give me a little bit more detail about what submodule you are interested or about the performance metrics you are hoping to add?

Best,

Dave

[Monika Shah]

I appreciate your prompt reply.

I would like to find time, space, data proximity, power for different components(like per tuple cost, per memory page cost, per  IO channel cost) used by each query plan.

[David Storch]

Hi Monika,

I would recommend reading the code path for the explain command. Explain calls into the Command::explain() virtual method, implemented by CRUD commands such as find and aggregate. The entry point for explain of a find command, for example, is here:

https://github.com/mongodb/mongo/blob/edc3dfce51b00ba3007f35cf6383efb72bf8b74e/src/mongo/db/commands/find_cmd.cpp#L139

For the aggregate command, the entry point is here:

https://github.com/mongodb/mongo/blob/edc3dfce51b00ba3007f35cf6383efb72bf8b74e/src/mongo/db/commands/pipeline_command.cpp#L99

You may also wish to read some of the explain-specific logic which lives in explain.{h,cpp}:

https://github.com/mongodb/mongo/blob/edc3dfce51b00ba3007f35cf6383efb72bf8b74e/src/mongo/db/query/explain.h

Explain currently exposes query-level execution statistics, not storage-level statistics. Collecting some of the metrics you mentioned may require some deeper collaboration with the storage layer. However, I hope that the pointers I've listed above are a good starting point.

Best,

Dave

[Monika Shah]

Link you shared is good to have some basic idea.
Can you share me code of query optimization?

[Monika Shah]

I have set break point find_cmd.cpp:132 (first line of find_cmd.cpp::explain() ), and find_cmd.cpp:231.
After that when I debug code, it stop at this line only for db.collection.findOne()  not for db.collection.find( predicate1, predicate2).count().


Please help me to see how does it verify criteria to prepare plan



















0