Pihole api unresponsive after consecutive requests

[lovelaze]

Versions

Core version is v6.0.4 (Latest: v6.0.4)
Web version is v6.0.1 (Latest: v6.0.1)
FTL version is v6.0.2 (Latest: v6.0.2)

Platform

• OS and version: Debian 12
• Platform: Docker

Expected behavior

The pihole api to not become unresponsive and refuse connections temporarily.

Actual behavior / bug

It seems like the pihole api becomes unresponsive briefly after sending the following request: PATCH /api/config.

A consecutive request to POST /api/action/gravity after a receving a http 200 from PATCH /api/config fails:

go http client error: http://192.168.0.148/api/action/gravity: Post \"http://192.168.0.148/api/action/gravity\": dial tcp 192.168.0.148:80: connect: connection refused"

This occurs intermittently, it's not unresponsive every time. Is this expected behavior?

Steps to reproduce

Steps to reproduce the behavior:

1. PATCH /api/config
2. POST /api/action/gravity
3. See error

Debug Token

• URL: https://tricorder.pi-hole.net/DSnCtToI/

[yubiuser]

Is this expected behavior?

Likely. There are many configuration changes, that require a restart before they have any effect. FTL is smart enough to know which changes require a restart and will automatically restart itself. During the restart, the API won't respond.

When you go to Settings > All Settings you see a small restart symbol for all options that require/trigger a restart

[lovelaze]

Likely. There are many configuration changes, that require a restart before they have any effect. FTL is smart enough to know which changes require a restart and will automatically restart itself. During the restart, the API won't respond.

Thanks, that makes sense. Do you have any recommendation of how to make consecutive api calls? I would like to avoid sleeping before proceeding with the next call, is there any preferred endpoint to call to "healthcheck" the api?

[yubiuser]

I don't think we have any specific health check endpoint. I myself would use /auth because it will also tell you if re-authentication is required.

[DL6ER]

I concur on using /auth for checking if the API is reachable. Alternatively, you can sleep only if changing an endpoint which is known to need a full restart. Use the GET parameter ?detailed=true when querying /config to get a boolean next to each of the config options telling you if a full restart is required or not. If not, there will be no downtime at all as FTL will be able to change the setting immediately at runtime.