Profiler
Suggest editsEDB SPL Check contains a simple profiler of SPL functions and procedures. It can work with or without access to shared memory and depends on the shared_preload_libraries
config.
When EDB SPL Check is initialized by shared_preload_libraries
, it can allocate shared memory and stores function profiles. When EDB SPL Check can't allocate shared memory, the profile is stored in session memory.
Due to dependencies, shared_preload_libraries
must contain edb-spl
first:
The profiler becomes active when the GUC spl_check.profiler
is on. The profiler doesn't require shared memory. However, if there's not enough shared memory, then the profiler is limited to an active session.
Activate the profiler by calling the function spl_check_profiler(true)
. Deactivate the profiler by calling the same function but with a false
argument (or off
with literals).
When shared_preload_libraries
initializes EDB SPL Check, the GUC spl_check.profiler_max_shared_chunks
becomes available to configure the amount of shared memory used by the profiler. This GUC defines the maximum number of statement chunks that can be stored in the shared memory. For each EDB SPL Check function/procedure, the whole content is split into chunks of 30 statements. If needed, you can use multiple chunks to store the whole content of a single function. A single chunk is 1704 bytes.
The default value for spl_check.profiler_max_shared_chunks
is 15000. This number is usually enough for big projects that contain hundred of thousands of statements and consume about 24MB of memory.
If your project requires fewer statement chunks, you can set the value for spl_check.profiler_max_shared_chunks
to a smaller number to decrease the memory usage.
The minimum value is 50, which usually consumes about 83kB of memory. The maximum value is 100000, which usually consumes about 163MB of memory.
Note
Changing spl_check.profiler_max_shared_chunks
requires a PostgreSQL restart.
The profiler also retrieves the query identifier for each instruction that contains an expression or optimizable statement. However, pg_stat_statements
(or another, similar third-party extension) must be installed.
The following are limitations of the query identifier retrieval:
- If a `EDB SPL Check expression contains underlying statements, only the top-level query identifier is retrieved.
- The profiler doesn't compute the query identifier and instead relies on the external extension, like
pg_stat_statements
. Depending on the external extension's behavior, a query identifier might not show for some statements. For example,pg_stat_statements
doesn't show the query identifier for queries with DDL statements. - The profiler retrieves a query identifier only for instructions containing expressions. So, the function
spl_check.profiler_max_shared_chunks
can report fewer query identifiers than instructions on a single line.
Updates on shared profiles can decrease the performance on servers under higher loads.
Display the profile with spl_profiler_function_tb
:
Display the profile per statements with spl_profiler_function_statements_tb
:
Display all stored profiles with spl_profiler_functions_all
:
Coverage metrics
EDB SPL Check provides two functions:
spl_coverage_statements(name)
spl_coverage_branches(name)
- On this page
- Coverage metrics
Could this page be better? Report a problem or suggest an addition!