Docs

Torch Log Service

EFS log collection

Collect EFS logs

This topic provides an introduction to EFS log features and collection methods.

Features overview

Elastic File Storage (EFS) is an AI cloud-native file storage solution offered by Volcano Engine, featuring elastic scaling, exceptional performance, and reliable security. It is suitable for AI model training, inference, autonomous driving, and enterprise office sharing, among other scenarios.

Log analysis can be enabled for EFS data migration policies or tasks. After enabling, you can query and analyze the data migration records of the policy in real time.

The related configuration is as follows:

Configuration	Description
Log type	EFS data movement log.
Log storage location	After enabling log analysis when creating a data movement policy or task, the corresponding project and log topic will be automatically created in TLS. For log resource management details, see Manage EFS log resources. Project: auditlog_efs_default Log topic: auditlog_efs_dataflow
Log storage period	It is defaulted to 3650 days.
Log shards	One shard is preset. The auto split feature for log topic shards is enabled by default.
Index settings	Configure indexes are automatically configured.
Log fields	For a detailed list of log fields, see EFS log fields.

Collect EFS logs

For information on usage restrictions and billing, see Data migration task overview.

For details on enabling log analysis and log collection, see Enable log analysis.

EFS log fields

Field	Description
cache_strategy_id	DataFlow policy ID.
file_name	Relative file path. For example, if the file system path in the DataFlow policy is /dir1/ and the absolute file path is /dir1/subdir1/file1, then the value here is subdir1/file1.
file_system_id	File system ID.
job_id	Task ID.
object_name	Object name.
size_bytes	File size, measured in bytes.
success	The status of the file synchronization result. The value descriptions are as follows: 1: Success. 0: Failure.
type	The task type of DataFlow task. The value descriptions are as follows: LoadMeta Event: Imports metadata immediately after the DataFlow policy is created. LoadData Event: Imports data immediately after the DataFlow policy is created. TosToNas Event: Imports incremental metadata and data. NasToToS Event: Exports data. AutoFileEvict Event: Automatically removes data. ManualFileEvict Event: Manually removes data.
skipped	Policy for handling files with the same name during import or export tasks: true: Skip the files with same names. false: Do not skip the files with same names.

Last updated: 2026.08.21 17:59:29