• Get an overview of cluster load and performance
• Drill down into your queries and understand the load they put on your cluster
• Search through logs and errors
• Monitor and kill running queries with the click of a button
• Get stats on your disk usage per node, and understand how much disk space tables, columns, and parts take up
• Run your own queries straight from the interface to further dig into performance and cluster issues
• Setup operations to run in the background with automatic rollbacks for failures

To deploy HouseWatch, clone this repo and then run the following, substituting the environment variables for the relevant values of one of your ClickHouse instances:

SITE_ADDRESS=<SITE_ADDRESS> \
CLICKHOUSE_HOST=localhost \
CLICKHOUSE_CLUSTER=mycluster \
CLICKHOUSE_USER=default \
CLICKHOUSE_PASSWORD=xxxxxxxxxxx \
docker compose -f docker-compose.yml up

SITE_ADDRESS here is the address that the UI will be running on. It can be a domain name or simply a port like :80.

After running the above, the UI will be running on the address you specified. This will be something like http://localhost if you used :80 for your SITE_ADDRESS above. I would think twice about exposing this to the internet, as it is not currently secured in any way.

To run HouseWatch locally along with a local ClickHouse instance, execute:

docker compose -f docker-compose.dev.yml up -d

then go to http://localhost:8080

At PostHog we manage a few large ClickHouse clusters and found ourselves in need of a tool to monitor and manage these more easily.

ClickHouse is fantastic at introspection, providing a lot of metadata about the system in its system tables so that it can be easily queried. However, knowing exactly how to query and parse the available information can be a difficult task. Over the years at PostHog, we've developed great intuition for how to debug ClickHouse issues using ClickHouse, and HouseWatch is the compilation of this knowledge into a tool.

Beyond monitoring, we also built internal systems and processes for managing the clusters that spanned various platforms. We would use Grafana to look at metrics, SSH into nodes for running operations and using specialized tooling, query via Metabase to dig deeper into the data in the system tables and create dashboards, and then a combination of tools baked into the PostHog product for further debugging and streamlined operations such as our async migrations tool, and internal views for listing queries and analyzing their performance.

As a result, we felt it was appropriate to have these tools live in one place. Ultimately, our vision for HouseWatch is that it can both serve the purpose of a pganalyze for the ClickHouse ecosystem, while also including tooling for taking action on insights derived from the analysis.

HouseWatch is in its early days and we have a lot more features in mind that we'd like to build into it going forward. The code could also use some cleaning up :) As of right now, it is considered Beta software and you should exercise caution when using it in production.

One potential approach is to connect HouseWatch to ClickHouse using a read-only user. In this case, the cluster management features will not work (e.g. operations, query editor), but the analysis toolset will function normally.

HouseWatch was created and is maintained by PostHog and yakkomajuri.

Contributions are certainly welcome! However, if you'd like to build a new feature, please open up an issue first.

A public list of things we intend to do with HouseWatch in the near future.