Continuous profiling is now GA

Last updated:

Find Hot Functions in the Table

Find functions with large self or total contributions to the selected OpenTelemetry profile measurement.

Select Table to rank functions by their contribution to the selected measurement.

Profiling function table sorted by Self, with Function Name, Self, Total, and Search functions controls

Read the columns

ColumnWhat to inspect
Function NameThe function to investigate.
SelfThe value and percentage attributed directly to the function, excluding its callees.
TotalThe value and percentage attributed to the function together with its callees.

See Self and Total for why totals overlap. Read values in the selected profile type's units. The screenshot uses On-CPU and shows time values.

Find the largest contributors

  1. Select the profile type, workload, and time range in the explorer.
  2. Select Table.
  3. Sort by Self to bring the largest direct contributors to the top, as shown in the screenshot.
  4. Compare Total with Self to distinguish work in the function from work in its callees.
  5. Enter a function name in Search functions... to find a known function in the table.

In the On-CPU screenshot, compress/flate.(*compressor).deflate has 1.35 minutes of self time and 3.06 minutes of total time. The difference comes from its callees. For any profile type, nearly equal Self and Total values mean that most of the contribution belongs to the function itself.

Continue with call context

The table ranks functions without showing their callers. Use Flame Graph to inspect call paths, or Sandwich for callers and callees.

Use Search functions... to find table entries. Use Add filter to apply stack and frame filters that change which code you analyze.

Further Reading