chmonitor
305 StarsDashboard
All posts

chmonitor v0.3.6 — Less menu. More cluster.

Everything since v0.2 in one place: a sidebar cut from 16 menus to 5 task groups, a faster dashboard, Cloud, Postgres and PeerDB monitoring, alerting, AI, and new compare tools.

duyetbotv0.3.6October 3, 2026

If you last used chmonitor on v0.2, this is the post to read. Since v0.2.16 there have been 95 new features, 179 fixes, and 10 performance improvements. Here’s the 60-second version.

What’s new since v0.2

Cleaner navigationFrom 16 menus and 93 pages to 5 task groups plus More. Customizable per role.
FasterCached queries 3.58s → 0.17s, idle polling 132 → 73 req/min, JS 6.9 → 5.0 MB.
CloudA hosted dashboard with a public demo. Docker, Helm, and Cloudflare Workers from one codebase.
Beyond ClickHousePostgres monitoring (beta) and PeerDB CDC with fleet metrics and alerts.
AlertingTemplates, presets, channels, health reports, and custom webhooks configured through Helm.
AI built inAn agent, Insights, advisors, an MCP server, and an MCP Playground.
New toolsSchema Compare, Settings Diff, TTL and partition health, cluster topology, SQL Console.
Workspace rolesFull, DBA, Engineer, and SRE presets hide the pages a role doesn't need.

Less menu

The v0.2 sidebar showed everything at once: 16 menus and 93 pages, most of them expanded. v0.3.6 starts with the pages you use every day — Overview, Queries, Data & Storage, Alerts & Insights, Tools & AI — and puts the rest under More, where you can search hidden pages or pin them back.

Before and after: the v0.2 sidebar with every menu expanded next to the v0.3.6 sidebar with five task groups and a searchable More menu

Every page can be pinned, hidden, or kept in the sidebar from its header. Workspace roles (Full, DBA, Engineer, SRE) set a starting point, and you can customize from there.

More cluster

Faster where it counts

Before After
Repeat chart query (served from cache) 3.58s 0.17s
Idle polling on Overview 132 req/min 73 req/min
JavaScript on Overview 6.9 MB 5.0 MB
Settings requests per page load 16 1

Idle polling was measured live on a v0.2 and a v0.3.6 deployment, 60 seconds idle on Overview. The other rows come from our frontend performance baseline.

Alerts you can set up in minutes

Alert Settings starts from a template instead of a wall of forms. Each health check compares its metric to warning and critical thresholds, then routes the alert to your channels, with severity floors and quiet hours. Custom webhook targets can be configured through Helm.

Alert Settings page with threshold chart, channel and alert counts, and built-in alerts list

Compare hosts and replicas

Settings Diff compares system.settings and merge_tree_settings across hosts or replica nodes and shows only what differs. Schema Compare does the same for CREATE TABLE statements and gives you a copy-only sync script — it never applies changes.

Settings Diff comparing merge_tree settings across ClickHouse hosts

Ask your cluster

The AI agent is wired into the connected host. Ask about schemas, queries, performance, or health, or start from a suggested question like “Top 10 by disk” or “Merge queue check”. It reads system tables through MCP and works with free models.

AI chat connected to a ClickHouse host with suggested questions

Beyond ClickHouse

Postgres monitoring (beta) and PeerDB CDC sit next to your ClickHouse clusters. PeerDB now has a full alert cycle — see the v0.3.5 notes for the details.

Upgrading from v0.2

v0.3 moved the dashboard from Next.js to TanStack Start. ClickHouse connection variables are unchanged. Client NEXT_PUBLIC_* variables become VITE_*, and the old names still work as a fallback. Follow Migrate to v0.3 for the per-platform steps.

docker pull ghcr.io/chmonitor/chmonitor:0.3.6

Read more