Shipped · Observability

What jaegertracing/jaeger shipped

The public repository of Jaegertracing · github.com/jaegertracing/jaeger

Written by FoxPlug from the repository's own releases; not affiliated with Jaegertracing. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/jaegertracing/jaeger; Jaegertracing did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.

Get a weekly update like this for your product, free

Or Use it as a GitHub Action

See the update FoxPlug writes for jaeger, free

Week of November 2, 2026

What shipped

Why it matters

This week combines a release with a set of breaking changes to the MCP tools and AI query endpoint that callers and configurations must account for. It also adds a new span search tool and hardens query serving against panics and slow deduplication. Operators upgrading should read the release notes and check their configuration for removed options.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

Jaeger v2.23.0 is out. The MCP endpoint now runs without sessions, find_traces takes only a sql statement, search_traces is renamed to find_traces, and a new find_spans tool lands. Query handlers also recover from panics now.

Post for LinkedIn

Jaeger v2.23.0 has been released. Alongside it, several changes landed for people using the MCP tools and the AI query endpoint: find_traces now takes a single sql statement, search_traces is renamed to find_traces, a new find_spans tool is available, and the turn-scoped MCP endpoint no longer uses sessions, with the stateless option removed. Query gRPC and MCP handlers now recover from panics, span ID deduplication runs in linear time, and the Elasticsearch writer stores non-finite doubles as string tags instead of dropping spans. The release workflow also verifies published assets and images after it finishes.

Week of October 26, 2026

What shipped

Why it matters

This week fixes several correctness problems people hit in daily use: trace walks no longer loop on duplicate span IDs, truncated topology responses now report total_span_count, and deep filter nesting is rejected before decoding. The v3 HTTP API gained prettyPrint, and the remaining v1 trace callers moved to the v3 endpoint ahead of its removal. Release and CI work also reduces the chance of publishing builds from the wrong commit.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

This week: v3 HTTP API supports prettyPrint, trace topology walks are guarded against duplicate span IDs, get_trace_topology reports total_span_count, and the last v1 trace callers move to the v3 API. Release runs can now build a given ref and flag commits past the tag.

Post for LinkedIn

This week in Jaeger: the v3 HTTP API now honors prettyPrint with the same 4-space indent as v1. Trace topology and critical path walks are guarded against duplicate span IDs, and get_trace_topology now reports total_span_count so truncated responses are visible to callers. Filter nesting depth is checked before api_v3 query parameters are decoded. Elsewhere, the last v1 trace endpoint callers in the repository moved to the v3 API ahead of its removal, unused ClickHouse feature gates were removed, and manual release runs can now build a given ref and flag commits past the tag.

Week of October 19, 2026

What shipped

Why it matters

This week adds a SQL input path for trace search, both as a package and in the MCP search_traces tool, and puts an authenticator in front of the AI endpoints. On the ClickHouse side, trace searches are bounded and time bounds keep nanosecond precision, while the query API rejects invalid metrics parameters and overflowing durations.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

This week: a SQL query surface for trace search and SQL input in the MCP search_traces tool, ai.auth to gate the AI endpoints, bounded ClickHouse trace searches with nanosecond time bounds, regex anchors and case folding, and stricter validation for metrics parameters and duratio

Post for LinkedIn

This week in Jaeger: a SQL query surface that compiles to trace search filters, and the MCP search_traces tool now accepts a SQL statement. An ai.auth setting can put an authenticator in front of the AI endpoints, and the AI configuration and routes moved into jaegerai. ClickHouse searches are bounded for distributed tables and GetTraces time bounds keep nanosecond precision. Regex filters honor ^ and $ anchors and (?i) case folding. The legacy metrics API rejects non-positive lookback, step and ratePer, durations that overflow are rejected, and non-finite float attributes render as strings.

Week of October 12, 2026

What shipped

Why it matters

This week several correctness fixes landed in storage, tenancy and query paths that change behavior for existing deployments. The ClickHouse schema change is breaking: a backend with create_schema enabled must now declare its table engine. MCP users running query behind a proxy or load balancer get options that address the session and Host header failures.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

This week: a breaking ClickHouse change requiring an explicit table_engine block, new MCP options for load-balanced and proxied query deployments, phrase and fulltext filters in the memory store, and fixes to auth resolution, tenancy checks and deduplication.

Post for LinkedIn

What shipped in Jaeger this week. ClickHouse schema creation now requires an explicit table_engine block, a breaking change for backends with create_schema enabled. The MCP endpoint gained streamable_http transport options for multiple query replicas behind a load balancer without session affinity, plus a setting to disable localhost protection for proxied deployments. The memory store now evaluates phrase and fulltext search operators. Fixes include deterministic resolution of storage auth extension references, stricter rejection of empty tenant values on gRPC, dropping empty resource spans after deduplication, rejecting non-positive lookback on the dependencies endpoint, and stopping Cassandra duration search at the configured depth.

Week of October 5, 2026

What shipped

Why it matters

This week's release and the milestones behind it make storage behavior visible to API clients, so callers can see what a backend can search before issuing queries. Several fixes also close incorrect or unsafe behavior, including span search sorting, filter routing to SQL, and tracing data paths through remote storage and MCP tool calls. For people running Jaeger, the changes reduce surprising results and harden the storage and container layers.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

Release v2.22.0 is out. Storage search capabilities now reach api_v3 over gRPC and HTTP, ClickHouse routes structured filters into SQL WHERE, and ES/OpenSearch no longer tie-break span search by document id by default. Filter tests and PromQL escaping also improved.

Post for LinkedIn

Release v2.22.0 is out this week. On top of it, Jaeger now exposes storage search capabilities over both api_v3 gRPC and HTTP, and the ClickHouse reader routes structured query filters into SQL WHERE clauses rather than folding them into a legacy shape. The Elasticsearch and OpenSearch reader declares span search, sorting and pagination capabilities, and the span search document id tie-breaker is now off by default. We also landed fixes for remote storage exports, MCP tool call spans, service name escaping in PromQL matchers, index name escaping in delete requests, and non-root primary groups for published images.

Week of September 28, 2026

What shipped

Why it matters

Span ordering is now fully implemented across query, storage gRPC, and backends, allowing callers to retrieve spans in their preferred order. Elasticsearch span search (RFC 0016 M4) progresses toward completion with tie-breaking, and release automation removes manual steps from the release process.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

Span ordering ships in query, gRPC storage, and memory backend with RFC 0016 contract. Elasticsearch span search progresses (M4) with tie-breaking. Release automation drafts GitHub releases automatically.

Post for LinkedIn

This week adds caller-selected span ordering across the query service, gRPC storage layer, and backends, implementing RFC 0016's FindSpans contract with deterministic tie-breakers and order-bound continuation tokens. Elasticsearch span search implementation (RFC 0016 M4) now includes sort tie-breaking on backing index and document ID. Release automation improvements remove manual steps from the release process. Two breaking changes update Elasticsearch configuration structure and remote storage setup.

Week of September 21, 2026

What shipped

Why it matters

Span search moves from interface design into working storage backends, pagination establishes a shared format across readers, and dead-letter handling completes lossless pipeline support. Feature gates advance from experimental to stable as code matures, while infrastructure fixes address timeout handling and buffer serialization.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

Span search reaches memory backend, pagination tokens standardized, and dead-letter handling completes lossless pipelines. Structured filters graduate to beta, native trace summaries to stable.

Post for LinkedIn

Week 40 advances RFC 0016 Span Search with FindSpans in memory storage and complete API v3 handlers, while RFC 0014 pagination establishes shared token format across readers. Dead-letter connector completes RFC 0007 lossless pipeline support. Feature gates progress: structured filters to beta, native trace summaries to stable. Fixes improve timeout handling and serialization correctness.

Week of September 14, 2026

What shipped

Why it matters

V2.21.0 is released with breaking changes and new span search capabilities. The paginated reader refactoring unifies how results flow through query and storage layers. Elasticsearch attribute indexing as numbers enables range queries to work correctly instead of lexicographically.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

Jaeger v2.21.0 is out. Span search (RFC 0016) is underway, Elasticsearch attributes now index as numbers for range queries, and paginated results are generalized across all reader methods.

Post for LinkedIn

Jaeger v2.21.0 ships with breaking changes: v1 HTTP endpoints removed, AI config restructured, and unsupported ES configs rejected at startup. New capabilities include paginated reader results unified across all methods and span search (RFC 0016) implementation starting with RFC 0016 M2. Elasticsearch storage now indexes attribute values as numbers, fixing range queries that previously compared lexicographically.

Weeks with too little public activity are left out rather than filled in. Last updated 2026-11-09.

Is this your repo? Ask us to remove this page.