Core Concepts¶

Skippr Architecture¶

Skippr Architecture

At-least Once¶

Skippr guarantees at-least once delivery to the output. Checkpoints of ingested data from the input are tracked internally and committed once the data is synced to the output.

Exactly Once semantics are in development where the output plugins will commit offsets on sync.

Checkpointing¶

Each input plugin implements its own checkpointing method according to the semantics of the data source.

For data sources like Kafka, this is simply a case of implementing the offsets capability of the Kafka Consumer API.

A more complex example of checkpointing can be found in the S3 input plugin, which tracks the object store last modified data and file line ingested.

Formats, Serializers and Deserializers¶

Skippr supports various common data input and output formats. Universal (de)serialization is central to Skipprs connect anything to anything principle.

Skippr serde's are generally wrappers around common C libs with extra validation and formatting auto-fixing.

Supported Formats¶

Json
Delimiter (, ; | \t)
Avro
Parquet
Arrow (Coming Soon)

Auto-fixing¶

... what do you mean by that?

The lessons from hundreds of data sources, trillions of records and years of (frankly painful) data integration have been distilled into the auto-fixing serde's.

Whether it's unicode json, missing or incorrect quotes, syntax errors or magic-byte padding... Skippr will do it's level best to auto-fix on ingest avoiding dozens of common causes of broken pipelines and hours of wasted engineering hours.

Skippr State Storage¶

Skippr creates and maintains the state file skippr-state.json in your state backend (e.g. local docker /data volume). This file contains Skippr internal metadata such as data source schema and checkpoints for ingestion progress (think kafka offsets or logstash sincedb).

NOTE: no source data or secrets are stored in the state.

Skippr File Buffer¶

Skippr manages temporary file buffers while ingesting data. The default buffer backend is file located your local volume mount (e.g. ~/demo/buffer directory).