Allow setting the IP bans reset interval via configuration and remove duplicate execution of cronjob to clean bans

[josecelano]

Relates to: https://github.com/torrust/torrust-tracker/issues/1452

The tracker has a BanService to ban clients' IPs sending many requests with the wrong connection ID.

There are two tasks:

• Set the interval in a config value
• Remove duplicate execution of the cronjob to clean the bans

Context

The BanService is shared across all UDP tracker servers. If a client is banned on one UDP server, it will also be banned on other UDP tracker servers running on different ports.

Task1: Set config value

The problem with adding a congif option is there is no section in the current configuration for services that are shared between all UDP trackers.

The current (not complete) configuration with 2 UDP servers:

[metadata]
app = "torrust-tracker"
purpose = "configuration"
schema_version = "2.0.0"

[logging]
threshold = "info"

[core]
inactive_peer_cleanup_interval = 600
listed = false
private = false
tracker_usage_statistics = true

  [core.announce_policy]
  interval = 300
  interval_min = 300

  [core.database]
  driver = "sqlite3"
  path = "/var/lib/torrust/tracker/database/sqlite3.db"

  [core.net]
  external_ip = "0.0.0.0"
  on_reverse_proxy = true

  [core.tracker_policy]
  max_peer_timeout = 900
  persistent_torrent_completed_stat = true
  remove_peerless_torrents = true

[[udp_trackers]]
bind_address = "0.0.0.0:6868"
tracker_usage_statistics = false

  [udp_trackers.cookie_lifetime]
  secs = 3_600
  nanos = 0

[[udp_trackers]]
bind_address = "0.0.0.0:6969"
tracker_usage_statistics = false

  [udp_trackers.cookie_lifetime]
  secs = 120
  nanos = 0

[[http_trackers]]
bind_address = "0.0.0.0:7070"
tracker_usage_statistics = false

[http_api]
bind_address = "0.0.0.0:1212"

  [http_api.access_tokens]
  admin = "***"

[health_check_api]
bind_address = "127.0.0.1:1313"

I propose to add a new section, udp_tracker_server, with configuration for that package. The same way we have a [core] section for the core tracker.

The new section with the new option would be:

[udp_tracker_server]
ip_bans_reset_intervals_in_secs = 3600

The default value will be 3600 * 24.

Task2: Swpan only one task to clean the bans

Now, every time the tracker runs a new UDP server, it spawns a new task to reset the bans:

tokio::spawn(async move {
    let mut cleaner_interval = interval(Duration::from_secs(IP_BANS_RESET_INTERVAL_IN_SECS));

    cleaner_interval.tick().await;

    loop {
        cleaner_interval.tick().await;
        ban_cleaner.write().await.reset_bans();
    }
});

It works because we launch all the UDP servers simultaneously when the tacker starts, but this should be done only once at the main app bootstrapping when the tracker launches jobs.

I labeled it a bug because, in practice, the bans are being reset more often than once per hour. They are being reset as many times as the number of UDP servers.

Relates to: https://github.com/torrust/torrust-tracker/issues/1444

cc @da2ce7