An open-source tool to visually explore and monitor ClickHouse. Resource attribution, merge activity, query performance, and engine internals in one place.
Inspect queries, merges, and mutations at any point in time. See what the cluster was doing and when.
Real-time View
Watch your cluster workload as it happens. CPU, memory, disk I/O, and network attributed in real time. From here you can deep dive into queries, merges, or parts from any metric.
Database Explorer
Browse databases, tables, parts, and columns. Inspect part lineage, merge history, and storage efficiency.
Merge Tracker
Live merge monitoring with dependency diagrams, timeline views, and throughput analysis.
Query Monitor
Running queries with per-query resource attribution, query anatomy breakdowns, and historical analysis from query_log.
Engine Internals
An attempt to visualise the internals of the ClickHouse engine as described in ClickHouse – Lightning Fast Analytics for Everyone (VLDB 2024). Thread pools, memory allocators, CPU sampling, PK index efficiency, and dictionaries.
Single-file HTML that runs entirely in your browser with any CORS-enabled ClickHouse instance.
The online version will ask for your ClickHouse credentials. If you prefer not to enter them on a third-party page, download it from the releases page or build it yourself.
Frequently Asked Questions
Some views use 3D to show relationships between parts, merges, and resources that are hard to represent in flat charts. If you prefer a simpler view, every screen has a 2D mode.
TraceHouse only runs read-only queries against system tables. On connect it probes the server for available capabilities (log tables, introspection functions, profiler settings, cluster topology, Keeper, Cloud mode) and gracefully disables features that aren't available. Some features (e.g. query profiling, trace_log) may require enabling specific settings on your server. See Connecting to ClickHouse for details.
TraceHouse is developed against recent ClickHouse releases. Older versions may work but some features that depend on newer system tables will be unavailable. The capability detection on connect will tell you what's missing.
Yes. Connections are stateless (HTTP-based), so you can configure multiple ClickHouse connections and switch between them in the UI. Each cluster gets its own overview dashboard, merge tracker, and query monitor.
TraceHouse polls system tables at configurable intervals (default 5 s). We are actively working on optimising query overhead. You can check the actual impact yourself via the built-in App Internal Monitor, which tracks the tool's own CPU, memory, and network usage per request.
TraceHouse is a client-side application. It runs entirely in the browser and does not install anything on your ClickHouse server or Grafana backend. It can be deployed as a Grafana app plugin, reusing your existing Grafana ClickHouse datasource and permissions, or run independently as a standalone web app.
TraceHouse only performs read-only queries over the HTTP interface, but we strongly recommend using a dedicated read-only ClickHouse account. We strongly suggest testing it against a staging or development cluster first. Pay attention to query load, memory usage, and network overhead that matter for your environment. That said, we haven't tested it at every scale. If you run into issues with large clusters or high-cardinality workloads, we'd love to hear about it.