Uptime config doesn't survive Redis restarts

[aldy505]

Environment

self-hosted 25.2.0, commit cut off: 9574780

Steps to Reproduce

On a running self-hosted instance, just assume we've applied all these settings: getsentry/self-hosted#3304 (comment)

Then do another ./install.sh on the self-hosted repo, which basically runs (1) docker compose down, (2) execute some bash stuff to migrate database and all that, (3) docker compose up -d. The redis data is not persisted, therefore the uptime:configs:{id} keys on redis would disappear, making the uptime checks not working.

Expected Result

Uptime checks works and can survive Redis restarts.

Actual Result

No uptime checks was done. The uptime-results Kafka topic is empty.

To hack this, I disable and enable the uptime monitor from the web UI, and the uptime:configs:{id} keys on redis was repopulated. But this is kind of a burden if you have lots of monitors.

My question would be: How can fix this? Should we stay on Redis and create an additional Redis container that has persisted volumes, or should we use some other persistence for storing configs?

[wedamija]

Yeah, basically we need to have a persistent redis set up for use with uptime. I know that metric alerts also uses redis as a persistent store... so we might already have a persistent redis node available in self hosted.

[aldy505]

Yeah, basically we need to have a persistent redis set up for use with uptime. I know that metric alerts also uses redis as a persistent store... so we might already have a persistent redis node available in self hosted.

Sadly we don't. But this is good to know. Regarding metric alerts, do you happen to know which part of Sentry that does/manage this?

[aldy505]

@wedamija Following up, do you have any issues with Redis on production? I don't think it's feasible for self-hosted, since Redis is shared with other services (sentry, relay). Do you have any other plans for adding another config provider as seen here?

uptime-checker/src/app/config.rs

Lines 15 to 19 in f9495e8

	#[derive(Debug, Copy, Clone, PartialEq, Eq, Deserialize, Serialize)]
	#[serde(rename_all = "snake_case")]
	pub enum ConfigProviderMode {
	Redis,
	}

[aldy505]

@wedamija Circling back to this, I just hit this issue again today. On self-hosted 25.4.0 (cutoff date of 15 April 2025), Redis took so much memory. Can we at least make a utility command on the sentry repo to re-populate the uptime:configs:{id}? Is that something that you think is possible?