All Products
Search
Document Center

Simple Log Service:Stability and limits of OSS data shipping

Last Updated:Jun 04, 2026

OSS data shipping (new version) provides high availability, auto-scaling, and data consistency. Review these limits before configuring a shipping job.

Stability

Read from SLS

Item

Description

Availability

High availability with built-in retry logic.

If a job fails to read data from SLS, it retries at least 10 times. If all retries fail, the job reports an error and restarts.

Write to OSS

Item

Description

Concurrency

Shipping instances are created per shard and scale out automatically.

When source Logstore shards split, new instances start within seconds.

Data consistency

Shipping scales out using consumer groups to ensure data consistency. Offsets commit only after data reaches OSS, preventing data loss.

Monitoring and alerting

Item

Description

Monitoring and alerting

Track metrics such as latency and traffic in real time. Set up alerts for issues such as insufficient export instances or exceeded network quotas. Configure alert monitoring rules to monitor data shipping jobs of the new version for OSS.

Limits

Network

Item

Description

Network type

Uses the Alibaba Cloud internal network for fast, stable transfers.

Permission management

Item

Description

Authorization

Grant the required permissions to ship and access data in OSS. Authorization overview.

Server-side encryption

If server-side encryption is enabled, grant additional permissions to the RAM role. OSS configuration documentation.

Read traffic

Item

Description

Read traffic

SLS enforces upper limits on read traffic per project and per shard. For more information, see Data read and write.

If a limit is exceeded, split shards or request a quota increase. A job that fails due to an exceeded limit retries at least 10 times. If all retries fail, it reports an error and restarts.

Write to OSS

Item

Description

Concurrent instances

Equals the total shard count, including read/write and read-only shards.

Data shipping

  • Uncompressed data size per OSS object: 5 to 256 MB, based on log data size per shard.

  • Shipping interval between consecutive jobs per shard: 300 to 900 seconds.

  • Each shipping job per instance generates one OSS object.

Time partition

Each batch writes one OSS object. The object path uses the earliest receive_time (when SLS receives the data) in the batch.

File format

Supports CSV, JSON, Parquet, and ORC formats. For more information, seeJSON format, CSV format, Parquet format, and ORC format.

Compression method

Supported: Snappy, gzip, zstd, or none.

OSS bucket

  • The destination OSS bucket must have Write Once Read Many (WORM) disabled and reside in the same region as the SLS project. For more information about WORM, see Bucket-level retention policy (BucketWorm).

  • Supported storage classes: Standard, Infrequent Access (IA), Archive, and Cold Archive. Objects inherit the bucket's storage class by default. Overview.

  • IA, Archive, and Cold Archive classes enforce minimum storage duration and minimum billable size. Choose based on retention and access needs. Overview.

Configuration items

Item

Description

Shipping latency

The shipping latency must not exceed the Logstore data retention period.

Reserve a buffer. For example, if data retention is 30 days, set shipping latency to 25 days or less.

Data shipping management

Item

Description

Job pause and resume

A paused job records the cursor of the last shipped entry and resumes from that point:

  • If the pause is within the data retention period, shipping resumes without data loss.

  • If the pause exceeds the retention period, shipping resumes from the nearest available data point.