launcher/src/power.rs

//! `power on` / `power off` subcommands.
//!
//! Power-off: SSH `sudo shutdown now` to each node → wait
//! `[home_assistant].shutdown_delay_seconds` (default 30s) → POST
//! `turn_off` to Home Assistant for each per-node `entity` and each
//! cluster-level `extra_entities`.
//!
//! Power-on: just POST `turn_on` to Home Assistant for each per-node
//! `entity` and each cluster-level `extra_entities` (the smart plug
//! handles waking the machine).

use anyhow::Result;
use std::time::Duration;
use tracing::{info, warn};

use crate::home_assistant::split_entity;

use crate::config::{Cluster, Config, MikrotikConfig, Node};
use crate::home_assistant::Client as HaClient;

mod mikro;

/// Flow for `power off`: SSH shutdown + wait + HA `turn_off`.
pub async fn power_off(
    config: &Config,
    cluster: &Cluster,
    node_names: &[String],
    dry_run: bool,
) -> Result<()> {
    let targets = select_targets(cluster, node_names, "power off")?;

    if dry_run {
        warn!("{}", fluent_i18n::t!("power-off-dryrun"));
        mikro::dry_off(config, &targets, &extra_ids(cluster));
        return Ok(());
    }

    // 1) SSH shutdown every target in parallel.
    // NOTE: This is fire-and-forget — if a node is unreachable, its
    // SSH failure is logged but the flow continues.  The HA `turn_off`
    // still fires for all targets regardless of SSH success.  This is
    // by design: already-offline nodes get their HA entity toggled off,
    // which is idempotent.
    info!(
        "{}",
        fluent_i18n::t!("power-off-head", { "nodes" => targets.len() })
    );
    let mut handles = Vec::with_capacity(targets.len());
    for node in targets.iter().cloned() {
        let config = config.clone();
        handles.push(crate::locale::spawn_with_locale(async move {
            let cmd = "sudo shutdown now".to_string();
            let r = crate::ssh::run_ssh(&config, &node, cmd.clone(), false).await;
            (node, cmd, r)
        }));
    }

    let mut any_ssh_failed = false;
    for h in handles {
        let (node, cmd, r) = h.await?;
        if let Err(e) = r {
            any_ssh_failed = true;
            warn!(
                "{}",
                fluent_i18n::t!("power-off-shutdown-failed", { "name" => &node.name, "cmd" => &cmd, "error" => e.to_string() })
            );
        }
    }

    // 2) Wait so the OS actually shuts down before we cut power.
    let delay = config
        .home_assistant
        .as_ref()
        .map(|c| c.shutdown_delay_seconds)
        .unwrap_or(30);
    info!(
        "{}",
        fluent_i18n::t!("power-off-waiting", { "secs" => delay })
    );
    tokio::time::sleep(Duration::from_secs(delay)).await;

    // 3) HA turn_off: per-node entities, then cluster-level extras.
    //    Mikrotik names are skipped here — they get a graceful RouterOS
    //    API shutdown in step 4 (after HA, so the router dies last).
    let mikrotik_off: Vec<String> = if let Some(ha_cfg) = &config.home_assistant {
        let ha = HaClient::new(ha_cfg)?;
        crate::home_assistant::warn_if_token_looks_short(&ha_cfg.token);
        toggle_entities(
            &ha,
            "turn_off",
            config.mikrotik.as_ref(),
            &targets,
            &extra_ids(cluster),
            "cluster",
        )
        .await
    } else {
        info!("{}", fluent_i18n::t!("power-off-no-ha"));
        // No HA at all — still shut down any mikrotik device referenced.
        mikro::collect_off(config, &targets, &extra_ids(cluster))
    };

    // 4) RouterOS graceful shutdown for each referenced device.
    mikro::shutdown_named(config, &mikrotik_off).await;

    if any_ssh_failed {
        warn!("{}", fluent_i18n::t!("power-off-ssh-failures"));
    }
    Ok(())
}

/// Flow for `power on`: HA `turn_on` for each node with an `entity`
/// and each cluster-level `extra_entities`.  No SSH needed — the smart
/// plug handles waking the machine.
pub async fn power_on(
    config: &Config,
    cluster: &Cluster,
    node_names: &[String],
    dry_run: bool,
) -> Result<()> {
    let targets = select_targets(cluster, node_names, "power on")?;

    let Some(ha_cfg) = &config.home_assistant else {
        anyhow::bail!(
            "[home_assistant] section is required for `power on` \
             (the launcher has no other way to wake a node)"
        );
    };
    let ha = HaClient::new(ha_cfg)?;
    crate::home_assistant::warn_if_token_looks_short(&ha_cfg.token);

    if dry_run {
        warn!("{}", fluent_i18n::t!("power-on-dryrun"));
        mikro::dry_on(config, &targets, &extra_ids(cluster));
        return Ok(());
    }

    info!(
        "{}",
        fluent_i18n::t!("power-on-head", { "nodes" => targets.len() })
    );
    let _ = toggle_entities(
        &ha,
        "turn_on",
        config.mikrotik.as_ref(),
        &targets,
        &extra_ids(cluster),
        "cluster",
    )
    .await;
    Ok(())
}

/// Entity ids of a cluster's `extra_entities` (both forms).
fn extra_ids(cluster: &Cluster) -> Vec<String> {
    cluster
        .extra_entities
        .iter()
        .map(|e| e.entity().to_string())
        .collect()
}

/// Handle one entity id for a power toggle: dispatch `[mikrotik]`
/// devices to the RouterOS API path (or their HA relay), everything else
/// to HA with `domain.name` validation.
async fn toggle_one(
    ha: &HaClient,
    service: &'static str,
    mikrotik: Option<&MikrotikConfig>,
    entity: &str,
    name_for_log: &str,
    mikrotik_off: &mut Vec<String>,
) {
    // Mikrotik devices are addressed by exact name — dispatch before any
    // HA validation (a router name could otherwise parse as a fictional
    // `mikrotik.<x>` HA entity).
    if let Some(dev) = mikro::device(mikrotik, entity) {
        if service == "turn_off" {
            if !mikrotik_off.iter().any(|n| n == &dev.name) {
                mikrotik_off.push(dev.name.clone());
            }
            info!(
                "{}",
                fluent_i18n::t!("power-off-mikrotik-skip", { "entity" => entity.to_string(), "address" => dev.address.clone() })
            );
            return;
        }
        if let Some(relay) = &dev.on_entity {
            if let Err(e) = call(ha, relay, service).await {
                warn!(
                    "{}",
                    fluent_i18n::t!("ha-extra-failed", { "service" => service, "extra" => relay.clone(), "label" => name_for_log.to_string(), "error" => e.to_string() })
                );
            }
        } else {
            warn!(
                "{}",
                fluent_i18n::t!("power-on-mikrotik-manual", { "name" => dev.name.clone() })
            );
        }
        return;
    }
    if let Err(e) = split_entity(entity) {
        warn!(
            "{}",
            fluent_i18n::t!("ha-invalid-entity", { "name" => name_for_log.to_string(), "entity" => entity.to_string(), "error" => e.to_string() })
        );
        return;
    }
    if let Err(e) = call(ha, entity, service).await {
        warn!(
            "{}",
            fluent_i18n::t!("ha-call-failed", { "service" => service, "entity" => entity.to_string(), "error" => e.to_string() })
        );
    }
}

/// Call HA `service` for each per-node entity, then each cluster-level
/// `extra_entities`.  Entities that name a `[mikrotik]` device are NOT
/// sent to HA: for `turn_off` their device names are returned (the caller
/// does the RouterOS API shutdown afterwards); for `turn_on` an optional
/// `on_entity` relay is toggled in HA instead, or a manual-power warning
/// is logged.
async fn toggle_entities(
    ha: &HaClient,
    service: &'static str,
    mikrotik: Option<&MikrotikConfig>,
    targets: &[Node],
    extras: &[String],
    extras_label: &'static str,
) -> Vec<String> {
    let mut mikrotik_off = Vec::new();
    for node in targets {
        if let Some(entity) = &node.entity {
            toggle_one(ha, service, mikrotik, entity, &node.name, &mut mikrotik_off).await;
        } else {
            warn!(
                "{}",
                fluent_i18n::t!("ha-skip-no-entity", { "name" => &node.name })
            );
        }
    }
    for extra in extras {
        toggle_one(
            ha,
            service,
            mikrotik,
            extra,
            extras_label,
            &mut mikrotik_off,
        )
        .await;
    }
    mikrotik_off
}

/// Dispatch to the right HA service.  Inlined because each call only
/// differs by `service`, and inlining keeps the trait surface small.
async fn call(ha: &HaClient, entity: &str, service: &'static str) -> Result<()> {
    match service {
        "turn_off" => ha.turn_off(entity).await,
        "turn_on" => ha.turn_on(entity).await,
        other => anyhow::bail!("unsupported HA service '{}'", other),
    }
}

/// Flow for `power switch --from <src> --to <tgt>`:
/// 1. Shut down + HA turn_off nodes/entities unique to `from`,
/// 2. HA turn_on all nodes/entities in `to`.
pub async fn power_switch(
    config: &Config,
    from: &Cluster,
    to: &Cluster,
    dry_run: bool,
) -> Result<()> {
    let to_names: std::collections::HashSet<&str> =
        to.nodes.iter().map(|n| n.name.as_str()).collect();
    let disable_nodes: Vec<Node> = from
        .nodes
        .iter()
        .filter(|n| !to_names.contains(n.name.as_str()))
        .cloned()
        .collect();
    let to_extras: std::collections::HashSet<&str> =
        to.extra_entities.iter().map(|e| e.entity()).collect();
    let disable_extras: Vec<String> = from
        .extra_entities
        .iter()
        .filter(|e| !to_extras.contains(e.entity()))
        .map(|e| e.entity().to_string())
        .collect();

    info!(
        "{}",
        fluent_i18n::t!("power-switch-head", { "from" => &from.name, "to" => &to.name, "nodes" => disable_nodes.len(), "entities" => disable_extras.len() })
    );

    if dry_run {
        mikro::dry_switch_off(config, &disable_nodes, &disable_extras);
        mikro::dry_switch_on(config, &to.nodes, &extra_ids(to));
        return Ok(());
    }

    // SSH shutdown nodes not in target
    let mut handles = Vec::with_capacity(disable_nodes.len());
    for node in disable_nodes.iter().cloned() {
        let c = config.clone();
        handles.push(crate::locale::spawn_with_locale(async move {
            (
                node.name.clone(),
                crate::ssh::run_ssh(&c, &node, "sudo shutdown now".into(), false).await,
            )
        }));
    }
    for h in handles {
        let (name, r) = h.await?;
        if let Err(e) = r {
            warn!(
                "{}",
                fluent_i18n::t!("power-switch-shutdown-failed", { "name" => &name, "error" => e.to_string() })
            );
        }
    }

    let delay = config
        .home_assistant
        .as_ref()
        .map(|c| c.shutdown_delay_seconds)
        .unwrap_or(30);
    if !disable_nodes.is_empty() {
        info!(
            "{}",
            fluent_i18n::t!("power-switch-waiting", { "secs" => delay })
        );
        tokio::time::sleep(Duration::from_secs(delay)).await;
    } else {
        // No nodes to shut down — only extra_entities differ.  HA
        // turn_off / turn_on runs immediately without delay.
    }

    if let Some(ha_cfg) = &config.home_assistant {
        let ha = HaClient::new(ha_cfg)?;
        crate::home_assistant::warn_if_token_looks_short(&ha_cfg.token);

        // Disable phase: per-node entities + extras unique to `from`.
        // Routed through `toggle_entities` (which validates + logs failures)
        // instead of ad-hoc loops with `let _ =` (which swallowed errors)
        // and dead `if dry_run` branches (dry-run returned at the top).
        let mikrotik_off = toggle_entities(
            &ha,
            "turn_off",
            config.mikrotik.as_ref(),
            &disable_nodes,
            &disable_extras,
            "from",
        )
        .await;

        // RouterOS devices named above: graceful API shutdown (after HA).
        mikro::shutdown_named(config, &mikrotik_off).await;

        info!(
            "{}",
            fluent_i18n::t!("power-switch-powering-on", { "nodes" => to.nodes.len(), "entities" => to.extra_entities.len(), "to" => &to.name })
        );
        let _ = toggle_entities(
            &ha,
            "turn_on",
            config.mikrotik.as_ref(),
            &to.nodes,
            &extra_ids(to),
            "to",
        )
        .await;
    } else {
        info!("{}", fluent_i18n::t!("power-switch-no-ha"));
        // No HA, but mikrotik devices named in `from` still get the
        // graceful API shutdown.
        mikro::shutdown_named(
            config,
            &mikro::collect_off(config, &disable_nodes, &disable_extras),
        )
        .await;
    }
    Ok(())
}

fn select_targets(cluster: &Cluster, node_names: &[String], op: &str) -> Result<Vec<Node>> {
    let targets: Vec<Node> = if node_names.is_empty() {
        cluster.nodes.clone()
    } else {
        let mut found: Vec<Node> = Vec::new();
        let mut missing: Vec<String> = Vec::new();
        for nn in node_names {
            if let Some(node) = cluster.nodes.iter().find(|n| &n.name == nn) {
                found.push(node.clone());
            } else {
                missing.push(nn.clone());
            }
        }
        if !missing.is_empty() {
            anyhow::bail!(
                "{}: node(s) {} not found in cluster '{}'",
                op,
                missing.join(", "),
                cluster.name
            );
        }
        found
    };
    if targets.is_empty() {
        anyhow::bail!("{}: cluster '{}' has no nodes", op, cluster.name);
    }
    Ok(targets)
}

#[cfg(test)]
mod tests;