Task Executions Table

This page describes the task executions table to help you understand and troubleshoot your jobs in Upsolver.

The system information in Upsolver is designed to help you to monitor and troubleshoot your jobs, by providing internal insights. Jobs are divided into various tasks, with each task responsible for working with data, performing maintenance work, and more. This section describes the task execution table to help understand your jobs.

The task executions table enables you to monitor the execution of the tasks that run your jobs and maintain your tables. To monitor and troubleshoot your jobs, run the following query:

SELECT * FROM logs.tasks.task_executions;

The following three sections describe the task executions table:

1. Task execution records: This section includes a list of fields in the task_executions table. It includes the field name and data type, as well as a short description of how to interpret each value.

2. Stage names: Upsolver operations comprise multiple stages that execute tasks to complete a job. This section describes each of these stages that can be found in the stage_name field. This can help you to better understand the progress of your jobs and identify the status of each stage.

3. Task event types: Each stage is a logical grouping of one or more tasks. This section describes the types of tasks that can be executed, along with a description of the task event types found in the task_event_types field.

Task execution records

Each record within the task_executions table describes a task being executed. The following is a list of the fields for each task:

Field name	Data type	Description
cluster_name	string	The name of the cluster that processed this task.
cluster_id	string	The unique ID of the cluster that processed this task.
cloud_server_name	string	The ID of the cloud instance this job is running on.
stage_name	string	Describes the type of task being executed. For descriptions of the different stage names, see Stage names.
job_name	string	The name of the job that the task belongs to.
job_id	string	The unique ID of the job that the task belongs to.
task_name	string	The name of the task formatted as the job_id with a prefix or suffix descriptor attached.
task_start_time	timestamp	The start time of the window of data being processed. In a transformation job, this corresponds to the value of run_start_time(). The difference between the task_start_time and task_end_time corresponds to the RUN_INTERVAL configured within the job options for transformation jobs. For data ingestion jobs, this defaults to 1 minute.
task_end_time	timestamp	The end time of the window of data being processed. This corresponds to the value of run_end_time() within transformation jobs. The difference between the task_start_time and task_end_time corresponds to the RUN_INTERVAL configured within the job options for transformation jobs. For data ingestion jobs, this defaults to 1 minute.
shard	bigint	The shard number corresponding to this task.
total_shards	bigint	The total number of shards used to process the job for this execution. This corresponds to the value configured by the EXECUTION_PARALLELISM job option. If the value of EXECUTION_PARALLELISM is altered at any point, the total_shards for future tasks belonging to that job are updated to match.
task_start_processing_time	timestamp	The time the task started being processed.
task_end_processing_time	timestamp	The time the task finished being processed.
task_items_read	bigint	The total number of records read.
bytes_read	bigint	The total bytes ingested from the source data in its original form, including header information.
bytes_read_as_json	bigint	The total bytes ingested from the source data if it were in a JSON format. This is the number used to determine the volume of data scanned for billing purposes.
duration	bigint	The time in milliseconds it took to process this task. This is equivalent to the difference between the task_start_processing_time and task_end_processing_time.
task_delay_from_start	bigint	The delay in milliseconds between the end of the data window and when the task began processing. This is equivalent to the difference between the task_end_time and task_start_processing_time.
task_classification	string	The classification of the task as user, system, input, or metadata based on the type of task being executed.
task_error_message	string	The error message, if an error is encountered.
task_event_type	string	Classifies the task into event types. For descriptions of the different event types, see Task event types.
organization_name	string	The name of the organization that the task belongs to.
log_processing_time	timestamp	The time the log record was processed.
organization_id	string	The unique ID of your organization (the same as the organization name).
partition_date_str	string	The partition date as a string.
partition_date	date	The date column that the table is partitioned by. Always qualify a partition_date filter in your queries to avoid full scans.
upsolver_schema_version	bigint	The system table's schema version. It changes when the user edits the output job that is written to this table.

Stage names

This section describes each of these stages that can be found in the stage_name field.

Stage name	Description
file discovery	Discovers the files within a file-based data source such as Amazon S3, Azure Blob Storage, or Google Cloud Storage.
data ingestion	Pulls data from the data source.
parse data	Parses the data discovered during file discovery or data ingestion stage.
Ingestion state maintenance	Performs maintenance work when data is being ingested.
write to storage	Writes output to object store.
write to target	Writes the data to the target location.
cleanup	Deletes old files that are unnecessary. This can be cleaning up unneeded files after compaction or removing other temporary files such as deleting batch files once the data has been parsed.
table state maintenance	Collects and maintains metadata about files as they are written to tables. This metadata is later used to perform tasks such as maintaining the file system, running compactions, running queries, and more.
retention	Deletes old data and metadata that have passed the retention period as configured when the table was created.
build indices	Builds indices for materialized views by reading the raw data and creating small files for the data that are then compacted and merged together.
compact indices	Compacts indices for materialized views after they have been built.
aggregation	Builds and compacts indices to perform aggregation for aggregated outputs.
collect statistics	Gathers metadata from the ingestion or output job by generating indexes.
compact statistics	Compacts and merges the metadata index.
partition metadata	Processes metadata for partition management and maintenance.
partition maintenance	Creates new partitions and deletes old ones.
partition management	Creates new partitions and deletes old ones.
count distinct metadata	Collects the number of distinct values for a field.
event type metadata	Builds the metadata index for a field when an event type is set in Upsolver Classic. This allows us to filter by event type and show statistics per event type.
upsert metadata	Maintains metadata about primary keys in order to know how and where to perform updates when they arrive as events.
monitoring metadata	Ensures metadata is written successfully.
dedup index	Builds the deduplication index. This index is used to run IS_DUPLICATE calculations.
coordinate compaction	Coordinates partition compactions by checking available files. Simultaneously maintains other table metadata.
compaction	Compacts smaller files into larger ones to optimize query performance when writing to a data lake output.
upsert compaction	Compacts data from multiple files to delete old rows that have a newer update.
compaction state maintenance	Performs maintenance work to ensure compaction state is healthy.
maintenance	Performs general maintenance tasks.
internal task	Performs tasks for working with connections to external environments.

Task event types

The following table describes the types of tasks that can be executed, along with a description of the task event types found in the task_event_types field.

Event type	Description
started	The task has started.
finished	The task has completed successfully .
heartbeat	An indicator that the task is still running. This is sent every 5 minutes to determine if a task is long-running and the current state of the task (so it has the current duration, read bytes and etc).
canceled	The task was canceled.
no-resources	Indicates a lack of resources to start a task. This is usually due to a connection limitation.
failed	The task has failed. Check task_error_message to better understand the error encountered.
failed-build	Failed to build a task.
failed-recoverable	An intermittent error has occurred (e.g. reading a file that was modified while reading it). The task will retry and recover from the error and the resulting data will be consistent.
dry-run-failed	The task from Upsolver's automated testing process; the testing of a new version has failed.
ignored-dry-run-failure	The dry run is ignored due to false positives.