fix(healthcheck): prevent ClickHouse broken-pipe errors in the telemetrystore healthcheck - #194

Fixes

• Stop the ClickHouse healthcheck from writing broken-pipe errors into the ClickHouse log.

Problem

The ClickHouse healthcheck uses wget --spider. The --spider option closes the connection before it reads the response body. ClickHouse then fails to write the /ping response, and records the failure as an Error with a full C++ stack trace.

The healthcheck runs every 30 seconds, and each failure writes two error lines. This adds about 5,700 stack traces a day. The container stays healthy the whole time, because the check still exits 0, so nothing surfaces the problem. The noise hides real ClickHouse errors.

This is the cause of the long-standing report in SigNoz/signoz#5140 (open since June 2024). A user identified the healthcheck there in September 2025, and a maintainer asked for a PR. This is that PR.

Fix

Replace --spider with -O-, so wget reads the response body.

Measurements

On clickhouse-server:25.5.6, running each command 25 times and counting new StaticRequestHandler error lines:

Verified on a live single-node deployment: the error count went from a steady stream to zero, and the container stays healthy with FailingStreak=0.

Scope

Only the coolify and ecsterraform castings are affected, because they use the 0.0.0.0 host. The dockercompose and dockerswarm castings use http://localhost:8123/ping and are left unchanged.

One thing worth flagging

Please do not fix this by changing the host to localhost. In the ClickHouse image, /etc/hosts resolves localhost to ::1 first. On a host with IPv6 disabled, the connection is refused:

$ wget -q -O- http://localhost:8123/ping
wget: can't connect to remote host: Connection refused     # rc=1

$ wget -q -O- 0.0.0.0:8123/ping
Ok.                                                        # rc=0

That variant produces no broken-pipe errors only because the request never reaches ClickHouse, and it would make the healthcheck fail. It may be worth a separate look at whether the dockercompose and dockerswarm castings are exposed to this on IPv6-disabled hosts. I have not tested those flavors, so I left them alone.

Refs SigNoz/signoz#5140