Performance instrumentation¶

MCP Client metrics¶

• mcpclient_job_total (Counter): number of jobs processed, labeled by script.
• mcpclient_job_success_timestamp (Gauge): timestamp of most recent job processed, labeled by script.
• mcpclient_job_error_total (Counter): number of failures processing jobs, labeled by script.
• mcpclient_job_error_timestamp (Gauge): timestamp of most recent job failure, labeled by script.
• mcpclient_task_execution_time_seconds (Histogram): histogram of worker task execution times in seconds, labeled by script.
• mcpclient_gearman_sleep_time_seconds (Counter): total worker sleep after gearman error times in seconds.
• mcpclient_transfer_started_total (Counter): number of Transfers started, by transfer type.
• mcpclient_transfer_started_timestamp (Gauge): timestamp of most recent transfer started, by transfer type.
• mcpclient_transfer_completed_total (Counter): number of Transfers completed, by transfer type.
• mcpclient_transfer_completed_timestamp (Gauge): timestamp of most recent transfer completed, by transfer type.
• mcpclient_transfer_error_total (Counter): number of transfer failures, by transfer type, error type.
• mcpclient_transfer_error_timestamp (Gauge): timestamp of most recent transfer failure, by transfer type, error type.
• mcpclient_transfer_files (Histogram): histogram of number of files included in transfers, by transfer type.
• mcpclient_transfer_size_bytes (Histogram): histogram of number of bytes in transfers, by transfer type.
• mcpclient_sip_started_total (Counter): number of SIPs started.
• mcpclient_sip_started_timestamp (Gauge): timestamp of most recent SIP started.
• mcpclient_sip_error_total (Counter): number of SIP failures, by error type.
• mcpclient_sip_error_timestamp (Gauge): timestamp of most recent SIP failure, by error type.
• mcpclient_aips_stored_total (Counter): number of AIPs stored.
• mcpclient_aips_stored_timestamp (Gauge): timestamp of most recent AIP stored.
• mcpclient_dips_stored_total (Counter): number of DIPs stored.
• mcpclient_dips_stored_timestamp (Gauge): timestamp of most recent DIP stored.
• mcpclient_aip_processing_seconds (Histogram): histogram of AIP processing time, from first file recorded in DB to storage in SS.
• mcpclient_dip_processing_seconds (Histogram): histogram of DIP processing time, from first file recorded in DB to storage in SS.
• mcpclient_aip_files_stored (Histogram): histogram of number of files stored in AIPs. Note, this includes metadata, derivatives, etc.
• mcpclient_dip_files_stored (Histogram): histogram of number of files stored in DIPs.
• mcpclient_aip_size_bytes (Histogram): histogram of number of bytes stored in AIPs. Note, this includes metadata, derivatives, etc.
• mcpclient_dip_size_bytes (Histogram): histogram of number of bytes stored in DIPs. Note, this includes metadata, derivatives, etc.
• mcpclient_aip_files_stored_by_file_group_and_format_total (Counter): number of original files stored in AIPs labeled by file group, format name. Note: format labels are intentionally not zeroed, so be aware of that when querying.
• mcpclient_aip_original_file_timestamps (Histogram): histogram of modification times for files stored in AIPs, bucketed by year.
• archivematica_info (Info): Archivematica version information.
• environment_info (Info): environment variable (configuration) information.

MCP Server metrics¶

• mcpserver_gearman_active_jobs (Gauge): Number of gearman jobs currently being processed.
• mcpserver_gearman_pending_jobs (Gauge): Number of gearman jobs pending submission.
• mcpserver_task_total (Counter): Number of tasks processed, labeled by task group, task name.
• mcpserver_task_error_total (Counter): Number of failures processing tasks, labeled by task group, task name.
• mcpserver_task_success_timestamp (Gauge): Most recent successfully processed task, labeled by task group, task name.
• mcpserver_task_error_timestamp (Gauge): Most recent failure when processing a task, labeled by task group, task name.
• mcpserver_task_duration_seconds (Histogram): Histogram of task processing durations in seconds, labeled by task group, task name, script name.
• archivematica_info (Info): Archivematica version information.
• environment_info (Info): Environment variable (configuration) information.

Dashboard metrics¶

The dashboard includes a large number of metrics provided by the django-prometheus package. These break down the number of HTTP requests, as well as request size, latency, errors, etc in a number of ways. They are primarily useful for developers working on the dashboard, although a number (e.g. django_http_requests_latency_including_middlewares_seconds for request latency or django_http_responses_total_by_status for number of HTTP responses) are useful for general reporting.

• django_http_responses_before_middlewares_total (Counter): Total count of requests before middlewares run
• django_http_requests_before_middlewares_total (Counter): Total count of responses before middlewares run
• django_http_requests_latency_including_middlewares_seconds (Histogram): Histogram of requests processing time (including middleware processing time)
• django_http_requests_unknown_latency_including_middlewares_total (Counter): Count of requests for which the latency was unknown (when computing * django_http_requests_latency_including_middlewares_seconds)
• django_http_requests_latency_seconds_by_view_method (Histogram): Histogram of request processing time labelled by view
• django_http_requests_unknown_latency_total (Counter): Count of requests for which the latency was unknown
• django_http_ajax_requests_total (Counter): Count of AJAX requests
• django_http_requests_total_by_method (Counter): Count of requests by method
• django_http_requests_total_by_transport (Counter): Count of requests by transport
• django_http_requests_total_by_view_transport_method (Counter): Count of requests by view, transport, method
• django_http_requests_body_total_bytes (Histogram): Histogram of requests by body size
• django_http_responses_total_by_templatename (Counter): Count of responses by template name
• django_http_responses_total_by_status (Counter): Count of responses by status
• django_http_responses_total_by_status_view_method (Counter): Count of responses by status, view, method
• django_http_responses_body_total_bytes (Histogram): Histogram of responses by body size
• django_http_responses_total_by_charset (Counter): Count of responses by charset
• django_http_responses_streaming_total (Counter): Count of streaming responses
• django_http_exceptions_total_by_type (Counter): Count of exceptions by object type
• django_http_exceptions_total_by_view (Counter): Count of exceptions by view
• django_model_save_total (Counter): Total model save calls labeled by model class
• django_model_delete_total (Counter): Total model save calls labeled by model class

The dashboard also includes additional metrics: common_db_retry_time_seconds (Counter): Total time waiting to retry database transactions in seconds, labeled by operation description common_ss_api_request_duration_seconds (Counter): Total time waiting on the Storage Service API in seconds, labeled by function

Storage Service metrics¶

The Storage Service includes all the same Django-based metrics as the Dashboard.

Configure Prometheus¶

Once Archivematica is configured to export variables, you can set up Prometheus to scrape data from it.

First, download and install Prometheus.

Create a Prometheus configuration file (in this example, it’s a YAML file called prometheus.yml). If you are running all Archivematica components on a single host, a good starting point is:

scrape_configs:
- job_name: archivematica
  scrape_interval: 5s
  static_configs:
  - targets:
          - localhost:7998
          - localhost:7999
          - localhost:62080
          - localhost:62081

Start Prometheus, passing in the path to the config file: ./prometheus --config.file=prometheus.yml.

You should be able to view the Prometheus web interface at http://<hostname>:9090. Under Status -> Targets in the menu bar, you’ll be able to see the status of the various Archivematica services.

Configure Grafana¶

Once you have Prometheus storing metrics, Grafana can be set up so that you can view them.

First, download and install Grafana.

After following the installation guide for your platform, you can login to the web interface (http://<hostname>:3000/, username: admin, password: admin). Note that the Grafana web interface can be exposed to the internet (after changing the default password), although you may wish to limit access.

Once you have logged into the Grafana interface, add Prometheus as a data source. After clicking Save & Test, you should see a “Data source is working” message.

Now you need to import a dashboard. There is a basic Archivematica dashboard available in Archivematica’s Docker compose environment, or you can download the dashboard JSON file directly from GitHub. In Grafana, under Dashboards -> Manage, choose Import, then either paste the JSON into the text box or save the dashboard as a file and upload it.