[SCT] Introduce the test for Manager installation and smoke checks

[mikliapko]

The idea is to have a very basic smoke test that verifies:

• Manager installation;
• possibility to run some very common operations.

The purpose is to run this test with non-main distros (debian10, debian11, ubuntu20) as the main idea of test executions on different platforms are mainly to verify Manager installation process.

Worth mentioning that we have never encountered the OS-specific Manager issues before which makes this testing approach viable.

[mikliapko]

@karol-kokoszka @Michal-Leszczynski I'd like to hear your recommendations here what to check in manager right after installation to verify installation process for different OS distros?

[karol-kokoszka]

Hey @mikliapko, the most important part of Scylla Manager server is the API server actually. The API server can be probed with /api/v1/version curl or by calling sctool version https://manager.docs.scylladb.com/stable/sctool/version.html. The latter is calling the same endpoint under the hood, so it validates if the API server is up and running.

In terms of scylla-manager-agent -> the same. The most important is API server to be up and running. Call /version endpoint.

Would be good to have some readiness/liveness endpoint available in manager to check if other servers are up and running too, like prometheus server and consul handler (@Michal-Leszczynski ).

[Michal-Leszczynski]

Would be good to have some readiness/liveness endpoint available in manager to check if other servers are up and running too, like prometheus server and consul handler (@Michal-Leszczynski ).

Do you mean that SM should query prometheus server which scrapes metrics from SM/agent and check if it's responding, or that SM should query SM/agent endpoints from which the metrics are being scraped and check if they are responding?

[karol-kokoszka]

Do you mean that SM should query prometheus server which scrapes metrics from SM/agent and check if it's responding, or that SM should query SM/agent endpoints from which the metrics are being scraped and check if they are responding?

We can discuss it on grooming.

[karol-kokoszka]

grooming notes

Current idea:

• install manager on the cluster + install agents
• it checks the availability of the SM http server curl --write-out "%{{http_code}}\n" --silent --output /dev/null "http://127.0.0.1:{port}/ping
• the same ping is needed on the Scylla Manager Agent's HTTP server
• sctool version which ping SM HTTP server