[sled-agent][package] Make zone uninstallation concurrent (#1178)

oxidecomputer · Jun 15, 2022 · 0b11803 · 0b11803
1 parent 84390f8
commit 0b11803
Show file tree

Hide file tree

Showing 2 changed files with 34 additions and 15 deletions.
diff --git a/package/src/bin/omicron-package.rs b/package/src/bin/omicron-package.rs
@@ -363,9 +363,10 @@ fn do_install(
 }
 
 fn uninstall_all_omicron_zones() -> Result<()> {
-    for zone in zone::Zones::get()? {
+    zone::Zones::get()?.into_par_iter().try_for_each(|zone| -> Result<()> {
         zone::Zones::halt_and_remove(zone.name())?;
-    }
+        Ok(())
+    })?;
     Ok(())
 }
 

diff --git a/sled-agent/src/sled_agent.rs b/sled-agent/src/sled_agent.rs
@@ -19,6 +19,7 @@ use crate::params::{
 };
 use crate::services::ServiceManager;
 use crate::storage_manager::StorageManager;
+use futures::stream::{self, StreamExt, TryStreamExt};
 use omicron_common::api::{
     internal::nexus::DiskRuntimeState, internal::nexus::InstanceRuntimeState,
     internal::nexus::UpdateArtifact,
@@ -169,10 +170,18 @@ impl SledAgent {
         // re-establish contact (i.e., if the Sled Agent crashed, but we wanted
         // to leave the running Zones intact).
         let zones = Zones::get()?;
-        for z in zones {
-            warn!(log, "Deleting existing zone"; "zone_name" => z.name());
-            Zones::halt_and_remove_logged(&log, z.name())?;
-        }
+        stream::iter(zones)
+            .zip(stream::iter(std::iter::repeat(log.clone())))
+            .map(Ok::<_, crate::zone::AdmError>)
+            .try_for_each_concurrent(
+                None,
+                |(zone, log)| async {
+                    tokio::task::spawn_blocking(move || {
+                        warn!(log, "Deleting existing zone"; "zone_name" => zone.name());
+                        Zones::halt_and_remove_logged(&log, zone.name())
+                    }).await.unwrap()
+                }
+            ).await?;
 
         // Identify all VNICs which should be managed by the Sled Agent.
         //
@@ -187,15 +196,24 @@ impl SledAgent {
         // Note that we don't currently delete the VNICs in any particular
         // order. That should be OK, since we're definitely deleting the guest
         // VNICs before the xde devices, which is the main constraint.
-        for vnic in Dladm::get_vnics()? {
-            warn!(
-              log,
-              "Deleting existing VNIC";
-                "vnic_name" => &vnic,
-                "vnic_kind" => ?VnicKind::from_name(&vnic).unwrap(),
-            );
-            Dladm::delete_vnic(&vnic)?;
-        }
+        let vnics = Dladm::get_vnics()?;
+        stream::iter(vnics)
+            .zip(stream::iter(std::iter::repeat(log.clone())))
+            .map(Ok::<_, crate::illumos::dladm::DeleteVnicError>)
+            .try_for_each_concurrent(None, |(vnic, log)| async {
+                tokio::task::spawn_blocking(move || {
+                    warn!(
+                      log,
+                      "Deleting existing VNIC";
+                        "vnic_name" => &vnic,
+                        "vnic_kind" => ?VnicKind::from_name(&vnic).unwrap(),
+                    );
+                    Dladm::delete_vnic(&vnic)
+                })
+                .await
+                .unwrap()
+            })
+            .await?;
 
         // Also delete any extant xde devices. These should also eventually be
         // recovered / tracked, to avoid interruption of any guests that are