gcloud alpha mldiagnostics machine-learning-run start-continuous-profiling

NAME
gcloud alpha mldiagnostics machine-learning-run start-continuous-profiling - start continuous profiling for a machine learning run
SYNOPSIS
gcloud alpha mldiagnostics machine-learning-run start-continuous-profiling --targets=[TARGETS,…] [--async] [--device-tracer-level=DEVICE_TRACER_LEVEL; default="enabled"] [--host-tracer-level=HOST_TRACER_LEVEL; default="info"] [--location=LOCATION] [--machine-learning-run=MACHINE_LEARNING_RUN] [--python-tracer-level=PYTHON_TRACER_LEVEL; default="disabled"] [--xprof-server-port=XPROF_SERVER_PORT] [GCLOUD_WIDE_FLAG …]
DESCRIPTION
(ALPHA) Start continuous profiling on one or more profiler targets of an active machine learning run.

While continuous profiling is active, the XProf profiler on each target keeps the most recent trace data in an in-memory circular buffer. To save the buffered data as a profiler session, run gcloud alpha mldiagnostics profiler-session capture with --kind=snapshot. On-demand profiler sessions can't be captured while continuous profiling is active.

Only one continuous profiling session can be active on a machine learning run at a time. The tracer levels and XProf server port can't be changed while continuous profiling is active. To change them, stop continuous profiling by running gcloud alpha mldiagnostics machine-learning-run stop-continuous-profiling, and then start it again.

EXAMPLES
To start continuous profiling on the profiler targets my-node-0 and my-node-1 of the machine learning run my-run in us-central1, run: 
gcloud alpha mldiagnostics machine-learning-run start-continuous-profiling --machine-learning-run=my-run --location=us-central1 --targets=my-node-0,my-node-1

To start continuous profiling with the lowest-overhead host tracer level and device tracing disabled, run: 

gcloud alpha mldiagnostics machine-learning-run start-continuous-profiling --machine-learning-run=my-run --location=us-central1 --targets=my-node-0 --host-tracer-level=critical --device-tracer-level=disabled

To list the profiler targets of the machine learning run my-run, run: 

gcloud alpha mldiagnostics profiler-target list --machine-learning-run=my-run --location=us-central1
REQUIRED FLAGS
ProfilerTarget resource - Profiler targets on which to start continuous profiling. To list the profiler targets of a machine learning run, run gcloud alpha mldiagnostics profiler-target list. This represents a Cloud resource. (NOTE) Some attributes are not given arguments in this group but can be set in other ways.

To set the project attribute:

  • provide the argument --targets on the command line with a fully specified name;
  • provide the argument --project on the command line;
  • set the property core/project.

To set the location attribute:

  • provide the argument --targets on the command line with a fully specified name;
  • provide the argument --location on the command line;
  • set the property compute/region.

To set the machine-learning-run attribute:

  • provide the argument --targets on the command line with a fully specified name;
  • provide the argument --machine-learning-run on the command line.

This must be specified.

--targets=[TARGETS,…]
IDs of the profilerTargets or fully qualified identifiers for the profilerTargets.

To set the profiler-target attribute:

  • provide the argument --targets on the command line.
OPTIONAL FLAGS
--async
Return immediately, without waiting for the operation in progress to complete.
--device-tracer-level=DEVICE_TRACER_LEVEL; default="enabled"
Device tracer level for the session. If the field is not set or unspecified, the default is DEVICE_TRACER_LEVEL_ENABLED. DEVICE_TRACER_LEVEL must be one of:
disabled
Tracer level is disabled.
enabled
Tracer level is enabled.
--host-tracer-level=HOST_TRACER_LEVEL; default="info"
Host tracer level for the session. If the field is not set or unspecified, the default is HOST_TRACER_LEVEL_INFO. HOST_TRACER_LEVEL must be one of:
critical
Tracer level is critical.
disabled
Tracer level is disabled.
info
Tracer level is info.
verbose
Tracer level is verbose.
--location=LOCATION
For resources [machine-learning-run, targets], provides fallback value for resource location attribute. When the resource's full URI path is not provided, location will fallback to this flag value.
--machine-learning-run=MACHINE_LEARNING_RUN
For resources [machine-learning-run, targets], provides fallback value for resource machine-learning-run attribute. When the resource's full URI path is not provided, machine-learning-run will fallback to this flag value.
--python-tracer-level=PYTHON_TRACER_LEVEL; default="disabled"
Python tracer level for the session. If the field is not set or unspecified, the default is PYTHON_TRACER_LEVEL_DISABLED. PYTHON_TRACER_LEVEL must be one of:
disabled
Tracer level is disabled.
enabled
Tracer level is enabled.
--xprof-server-port=XPROF_SERVER_PORT
Port of the XProf profiler server that runs in your workload on each profiler target. If unspecified, port 9999 is used.
GCLOUD WIDE FLAGS
These flags are available to all commands: --access-token-file, --account, --billing-project, --configuration, --flags-file, --flatten, --format, --help, --impersonate-service-account, --log-http, --project, --quiet, --trace-token, --user-output-enabled, --verbosity.

Run $ gcloud help for details.

API REFERENCE
This command uses the hypercomputecluster/v1alpha API. The full documentation for this API can be found at: https://docs.cloud.google.com/cluster-director/docs
NOTES
This command is currently in alpha and might change without notice. If this command fails with API permission errors despite specifying the correct project, you might be trying to access an API with an invitation-only early access allowlist.