Prometheus server.

[Yaa101]

Interesting project (for me) you have.
I was looking into your source code and saw that you use a Prometheus go client for instrumentation.
I was wondering which, or better said, who's Prometheus server do you interact with and what is the role of it within Litestream?

[yanc0]

Litestream exports metrics on -addr. Usually, applications don't interact with prometheus, it's prometheus that scrapes applications.

https://pkg.go.dev/github.com/prometheus/client_golang/prometheus

litestream replicate [arguments] DB_PATH REPLICA_URL [REPLICA_URL...]
Arguments:
[...]
        -addr BIND_ADDR
            Starts an HTTP server that reports prometheus metrics and provides
            an endpoint for live read replication. (e.g. ":9090")
[...]

[Yaa101]

Thanks for your answer, still it does not answer my quest to get the mental picture of what metrics go where and with what purpose.
I know that I am probably too lazy to read through the whole source tree to get the mental picture that way, I was hoping someone here could tell me some things about it, so I will have an easier time to comprehend the source, as it is vast.
My question is the cause of having seen the interaction graph on the Prometheus site, which states that the client reports (pushes) metrics to the Prometheus server, but I can't find inclusion references to the server in the source of Litestream.

[yanc0]

PushGateway is the only option to push metrics into Prometheus, and it's rarely used except for very special cases. Prometheus is pull model designed. It first discovers which application exposes metrics (using manual or auto discovery) and then, makes a GET HTTP request (usually on their /metrics endpoint) and then stores returned data on its internal TSDB.

I never seen clients (apps) reporting directly to Prometheus but I would be really happy if you have any reference from the official documentation saying the opposite.

Anyway, Litestream exposes some metrics like:

• checkpoints errors
• database informations
• wal informations
• etc

You can find the complete list by looking for promauto in the source code.

[Yaa101]

Thanks for explaining so far... questions hereunder are part retorical but if you want to help answering, your help is appreciated...
From this and reading documents on the Prometheus site I gather sofar that Litestream only uses the client_golang library to automatically register measurement instruments specific to Litestream and expose them through a http server.
It means that I could use the main Prometheus package to gather these metrics and do stuff with it?
Further, what is the Litestream specific usage of exposing these metrics, is it used by any end to do stuff with that?

[Yaa101]

I discovered the following in the configuration file section: https://litestream.io/reference/config/#metrics
"Litestream produces a continuous stream of metrics that are exported as a Prometheus endpoint. These are disabled by default as it requires an HTTP server to start. You can enable it by setting a bind address in the config"
So indeed I could look at the raw data at the configured http address and port or use a Prometheus system for that.
But there is no need to enable for Litestream to function properly.

I think this, with all the above explanation should answer my initial question, I learned a lot from this conversation to help me get the mental picture of how Litestream does what it is supposed to do.
I hope it helps other people too.
Thanks for participating.