From 596d82fe480248ce6c89b95990245f213145bba3 Mon Sep 17 00:00:00 2001
From: Thanos Makatos <thanos.makatos@nutanix.com>
Date: Mon, 6 Jul 2026 09:20:11 +0000
Subject: [PATCH] add optional dry run mode

This allows for all the decisions to be made and reported except the
actual decisions to be executed. This helps debugging.

Signed-off-by: Thanos Makatos <thanos.makatos@nutanix.com>
---
 io-thread-controller.json |  5 +++--
 src/config.rs             |  5 +++++
 src/controller.rs         | 25 +++++++++++++++++++++++++
 src/main.rs               |  6 ++++++
 4 files changed, 39 insertions(+), 2 deletions(-)

diff --git a/io-thread-controller.json b/io-thread-controller.json
index c6bb0f9..6f0dd44 100644
--- a/io-thread-controller.json
+++ b/io-thread-controller.json
@@ -8,6 +8,7 @@
   "cooldown_secs": 10.0,
   "engine": "threshold",
   "engine_config_dir": "/etc/io-thread-controller.d/engines",
-  "backend_config_dir": "/etc/io-thread-controller.d/backends"
-  "vm_state_path": "/run/io-thread-controller/vm-ownership.json"
+  "backend_config_dir": "/etc/io-thread-controller.d/backends",
+  "vm_state_path": "/run/io-thread-controller/vm-ownership.json",
+  "dry_run": false
 }
diff --git a/src/config.rs b/src/config.rs
index 58a4584..5f4f796 100644
--- a/src/config.rs
+++ b/src/config.rs
@@ -75,6 +75,10 @@ pub struct Config {
     /// Emit a one-time legend for status-line fields at startup.
     #[serde(default)]
     pub print_status_header: bool,
+    /// When true, log the scaling verdict but skip the actuation
+    /// call to `set_thread_count`.
+    #[serde(default)]
+    pub dry_run: bool,
 }
 
 fn default_true() -> bool {
@@ -127,6 +131,7 @@ impl Default for Config {
             enable_per_vm_status_line: true,
             enable_aggregate_status_line: true,
             print_status_header: false,
+            dry_run: false,
         }
     }
 }
diff --git a/src/controller.rs b/src/controller.rs
index 59e2557..09b64de 100644
--- a/src/controller.rs
+++ b/src/controller.rs
@@ -650,6 +650,31 @@ impl Controller {
             return Ok(());
         }
 
+        if self.cfg.dry_run {
+            tracing::info!(
+                target: "controller",
+                vm = %instance.id,
+                action = %action,
+                target,
+                "dry-run: would scale"
+            );
+            if matches!(action, ScaleAction::Up(_) | ScaleAction::Down(_)) {
+                instance.status.write().await.cooldown_until =
+                    Some(Instant::now() + Duration::from_secs_f64(self.cfg.cooldown_secs));
+            }
+            self.engine
+                .on_applied(
+                    &instance.id,
+                    AppliedOutcome::DryRun {
+                        action,
+                        prev_thread_count: previous_count,
+                        prev_io_count_total: previous_io_count,
+                    },
+                )
+                .await;
+            return Ok(());
+        }
+
         match instance.client.set_thread_count(target).await {
             Ok(()) => {
                 let mut status = instance.status.write().await;
diff --git a/src/main.rs b/src/main.rs
index dd0a159..4d6f959 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -77,6 +77,9 @@ struct Cli {
     /// Emit a one-time legend for uptime-style status fields.
     #[arg(long)]
     print_status_header: bool,
+    /// Log scaling decisions without calling set_thread_count.
+    #[arg(long)]
+    dry_run: bool,
 }
 
 /// See `Cli::log_style`.
@@ -123,6 +126,9 @@ async fn main() -> Result<(), IoThreadControllerError> {
     if cli.print_status_header {
         cfg.print_status_header = true;
     }
+    if cli.dry_run {
+        cfg.dry_run = true;
+    }
     validate_config(&cfg)?;
     let backends = registered_backends(&cfg)?;
     run(cfg, backends).await?;