Thank you for submitting this pull request. We really appreciate you spending the time to work on SurrealDB. 🚀 🎉

What is the motivation?

The implementation of distributed sequences in SurrealDB meets the critical need for generating reliable, monotonically increasing numeric identifiers across a cluster of nodes without excessive coordination overhead, allowing applications to utilize sequences for unique IDs, counters, and ordered operations with optimal performance in distributed environments.

What does this change do?

Distributed Sequence Generator for SurrealDB.

This code provides reliable, monotonically increasing numeric sequences in both single‑node and clustered SurrealDB deployments (multiple compute nodes backed by TiKV). It uses a batch‑allocation strategy to minimise coordination while guaranteeing global uniqueness.

Key features:

• Batch allocation: Nodes request ranges of sequence values at once, reducing network chatter and coordination overhead.
• Node ownership tagging: Every batch is tagged with the requesting node’s UUID to prevent overlap between nodes.
• Durable Persistence: Sequence metadata is stored in the underlying key-value store to survive restarts and network partitions.
• Concurrent, thread‑safe access: A DashMap caches active sequences, allowing lock‑free reads on the hot path.
• Exponential back‑off with full jitter: When a batch‑allocation attempt fails, the node retries with an exponential delay that includes full jitter to avoid thundering‑herd effects across the cluster.
• Automatic cleanup: Listens for namespace and database‑removal events and purges the corresponding sequence state.

How it works

The sequence implementation avoids contention by having each node reserve a range of sequence values, allowing it to serve multiple requests locally without requiring distributed coordination for every request. When a node exhausts its allocated range, it acquires a new batch from the distributed store.

To create a sequence:

DEFINE SEQUENCE mySeq;DEFINE SEQUENCE mySeq2 BATCH1000 START100 TIMEOUT 5s;

It is possible to modify the timeout of the sequence:

ALTERSEQUENCE mySeq TIMEOUT 5s;

Once a sequence is created, it is possible to retrieve the consecutive values:

RETURN sequence::nextval('mySeq');RETURN sequence::nextval('mySeq');

Will return:

01

To remove a sequence:

REMOVE SEQUENCE mySeq;

INFO FOR DB returns now defined sequences:

INFO FOR DB

{accesses:{},analyzers:{},apis:{},buckets:{},configs:{},functions:{},models:{},params:{},sequences:{seq1:'DEFINE SEQUENCE seq1 BATCH 1000 START 0',seq2:'DEFINE SEQUENCE seq2 BATCH 250 START -25',seq3:'DEFINE SEQUENCE seq3 BATCH 1000 START 0 TIMEOUT 5s'},tables:{},users:{}}

What is your testing strategy?

Github action.
Tests have been added to language tests and unit tests.

Is this related to any issues?

• No related issues

Does this change need documentation?

• Introduce sequences docs.surrealdb.com#1281

Have you read the Contributing Guidelines?

• I have read theContributing Guidelines