GreptimeDB v1.3: native histograms and Trace V2

Team GreptimeNov 2, 2026

Greptime Monthly · October 2026

v1.3 GA: native histograms and Trace V2

v1.3.0 is GA: native histograms are on by default, Trace V2 stores trace attributes in JSON2, and several incorrect SQL and PromQL results are fixed. If you run a cluster, read the upgrade note before you roll it out.

243

PRs merged

Community contributors

First-time

Release

v1.3.0

GA on October 28. The release notes cover automatic write retries, the PromQL compatibility gains, and every breaking change.

▸ Native histograms from Prometheus Remote Write v2 and OTLP are accepted by default, and PromQL can query them.

▸ SQL and PromQL fixes keep matching rows and requested ordering, retain IGNORE NULLS semantics, and stop NaN ratios from triggering false alerts. Fulltext bloom indexes no longer skip rows with underscores or mixed Han and ASCII text.

▸ The experimental semantic entity graph derives services, hosts, containers, and Kubernetes objects from metrics and traces at query time. information_schema.table_semantics shows what each table contributes.

Full release notes →

Before you upgrade to v1.3

▸ Upgrade all datanodes before frontends, and roll back frontends first. v1.3 frontends retry failed writes using per-entry results from datanodes; with old datanodes they can replay successful writes and duplicate rows in append-mode tables. max_retries = 0 under [write_retry] turns retries off.

▸ Keep metasrv maintenance mode on until the upgrade completes. On Kubernetes, use greptimedb-operator v1.3.0 or later.

▸ Review two changed defaults: scan_memory_limit is 25% of system memory, and VRL pipelines can’t call host functions unless pipeline.allow_vrl_host_access = true. Existing fulltext bloom indexes need a rebuild through compaction or ADMIN BUILD_INDEX.

All upgrade notes →

The idea behind v1.3

Observability Database: Should the Database Understand Telemetry?

The post defines an observability database as one that understands the semantics in telemetry data and uses them to optimize storage, query execution, and scaling. v1.3 applies this at write time: auto-created tables record metric.type, metric.unit, and metric.temporality, and metric.metadata_quality separates protocol-declared from suffix-inferred metadata. Read →

Feature spotlight · v1.3

Store trace attributes without a column per key

Trace V2 stores span, scope, and resource attributes in three JSON2 columns instead of one column per key. You pick the type at query time.

Opt in

greptime_trace_v2 in the pipeline header

Attributes go to JSON2

3 columns; span events and links kept

Query

Cast in SQL; Jaeger queries work too

OTel Collector exporter · SQL

# otlp_http exporter headers
x-greptime-pipeline-name: greptime_trace_v2
x-greptime-trace-table-name: traces_v2
-- quote attribute keys that contain dots
SELECT trace_id,
span_attributes."http.response.status_code"::BIGINT AS status
FROM traces_v2;

Opt-in, new table. Existing V1 tables are not migrated. Read the docs →

Featured

Use Case

3+ Billion Rows per Experiment: How a Commercial Space Company Moved from InfluxDB to GreptimeDB

Experiments of 3B+ rows ran InfluxDB 2.x out of memory. Why the team chose GreptimeDB over ClickHouse and VictoriaMetrics, and what replaced buckets. Read →

Engineering

Talk Recap | An Observability Database Across 7 Public Clouds: Where Data Lands and Where Queries Run

80+ clusters on 7 clouds, with log storage cost down more than 60% from Loki. Where OB Cloud keeps the data and where queries run. Read →

Benchmark

GreptimeDB v1.2 vs InfluxDB 3.11: Read/Write Performance and Storage Footprint

GreptimeDB has 1.31x the ingest throughput of InfluxDB 3 Enterprise and wins the all-host queries; Enterprise wins the 1- and 8-host ones. Numbers and repro steps inside. Read →

Upcoming Events

Upcoming

KubeCon + CloudNativeCon North America 2026

Nov 10–12, 2026 · Salt Lake City, UT

Greptime is a Silver Sponsor. Booth 262 has a live demo of OTLP metrics, logs, and traces in one GreptimeDB.

Meet Us at Booth 262

What's New

Oct 31

Dashboard Signals Explore, PostgreSQL COPY FROM STDIN, and everything else on main: Monthly Report No. 89 →

Community Contributors

14 community contributors merged 34 PRs this month.