Merge #1594

1594: Rebuild Rangers r=tiagolobocastro a=tiagolobocastro

    refactor(rebuild): use new rebuild rangers
    
    Make use of the rebuild rangers to configure rebuild types.
    This allows us to remove the setting of the rebuild map being done
    after the rebuild job is created for the nexus and removing it from
    the shared rebuild descriptor.
    
    The nexus still uses the partial but sequential rebuild to reduce
    the scope of changes.
    Once the fully partial rebuild is validated we can switch the nexus
    to it.
    
    Signed-off-by: Tiago Castro <tiagolobocastro@gmail.com>

---

    feat(rebuild): add rebuild rangers
    
    Adds a set of rebuild rangers which can be used by different rebuilds:
    1. full rebuild - walk the entire device range and copy every segment
       (current nexus full rebuild behaviour).
    2. partial rebuild - walk the allocated segments only and copy them.
    3. partial eq rebuild - walk the entire device range and copy only allocated
       segments (current nexus partial rebuild behaviour).
    
    Signed-off-by: Tiago Castro <tiagolobocastro@gmail.com>


Co-authored-by: Tiago Castro <tiagolobocastro@gmail.com>

io-engine/src/bdev/nexus/nexus_bdev_rebuild.rs

@@ -19,6 +19,7 @@ use crate::{
     rebuild::{
         HistoryRecord,
         NexusRebuildJob,
+        NexusRebuildJobStarter,
         RebuildError,
         RebuildJobOptions,
         RebuildState,
@@ -119,7 +120,8 @@ impl<'n> Nexus<'n> {
         }?;
 
         // Create a rebuild job for the child.
-        self.create_rebuild_job(&src_child_uri, &dst_child_uri)
+        let starter = self
+            .create_rebuild_job(&src_child_uri, &dst_child_uri)
             .await?;
 
         self.event(
@@ -146,8 +148,8 @@ impl<'n> Nexus<'n> {
             .lookup_child(&dst_child_uri)
             .and_then(|c| c.stop_io_log());
 
-        self.rebuild_job_mut(&dst_child_uri)?
-            .start(map)
+        starter
+            .start(self.rebuild_job_mut(&dst_child_uri)?, map)
             .await
             .context(nexus_err::RebuildOperation {
                 job: child_uri.to_owned(),
@@ -160,7 +162,7 @@ impl<'n> Nexus<'n> {
         &self,
         src_child_uri: &str,
         dst_child_uri: &str,
-    ) -> Result<(), Error> {
+    ) -> Result<NexusRebuildJobStarter, Error> {
         let verify_mode = match std::env::var("NEXUS_REBUILD_VERIFY")
             .unwrap_or_default()
             .as_str()
@@ -186,7 +188,7 @@ impl<'n> Nexus<'n> {
             verify_mode,
         };
 
-        NexusRebuildJob::new(
+        NexusRebuildJob::new_starter(
             &self.name,
             src_child_uri,
             dst_child_uri,
@@ -202,7 +204,7 @@ impl<'n> Nexus<'n> {
             },
         )
         .await
-        .and_then(NexusRebuildJob::store)
+        .and_then(NexusRebuildJobStarter::store)
         .context(nexus_err::CreateRebuild {
             child: dst_child_uri.to_owned(),
             name: self.name.clone(),

io-engine/src/core/segment_map.rs

@@ -96,4 +96,15 @@ impl SegmentMap {
     pub(crate) fn count_dirty_blks(&self) -> u64 {
         self.count_ones() * self.segment_size / self.block_len
     }
+
+    /// Get the segment size in blocks.
+    pub(crate) fn segment_size_blks(&self) -> u64 {
+        self.segment_size / self.block_len
+    }
+}
+
+impl From<SegmentMap> for BitVec {
+    fn from(value: SegmentMap) -> Self {
+        value.segments
+    }
 }

io-engine/src/rebuild/bdev_rebuild.rs

@@ -1,7 +1,4 @@
-use std::{
-    ops::{Deref, Range},
-    rc::Rc,
-};
+use std::ops::{Deref, Range};
 
 use super::{
     rebuild_descriptor::RebuildDescriptor,
@@ -13,7 +10,10 @@ use super::{
     SEGMENT_TASKS,
 };
 
-use crate::gen_rebuild_instances;
+use crate::{
+    gen_rebuild_instances,
+    rebuild::rebuilders::{FullRebuild, RangeRebuilder},
+};
 
 /// A Bdev rebuild job is responsible for managing a rebuild (copy) which reads
 /// from source_hdl and writes into destination_hdl from specified start to end.
@@ -59,47 +59,50 @@ gen_rebuild_instances!(BdevRebuildJob);
 /// A rebuild job which is responsible for rebuilding from
 /// source to target of the `RebuildDescriptor`.
 pub(super) struct BdevRebuildJobBackend {
-    /// The next block to be rebuilt.
-    next: u64,
     /// A pool of tasks which perform the actual data rebuild.
     task_pool: RebuildTasks,
     /// A generic rebuild descriptor.
-    descriptor: Rc<RebuildDescriptor>,
+    copier: FullRebuild<RebuildDescriptor>,
     /// Notification callback with src and dst uri's.
     notify_fn: fn(&str, &str) -> (),
 }
 
 #[async_trait::async_trait(?Send)]
 impl RebuildBackend for BdevRebuildJobBackend {
     fn on_state_change(&mut self) {
-        (self.notify_fn)(&self.descriptor.src_uri, &self.descriptor.dst_uri);
+        let desc = self.common_desc();
+        (self.notify_fn)(&desc.src_uri, &desc.dst_uri);
     }
 
     fn common_desc(&self) -> &RebuildDescriptor {
-        &self.descriptor
+        self.copier.desc()
+    }
+
+    fn blocks_remaining(&self) -> u64 {
+        self.copier.blocks_remaining()
+    }
+
+    fn is_partial(&self) -> bool {
+        self.copier.is_partial()
     }
 
     fn task_pool(&self) -> &RebuildTasks {
         &self.task_pool
     }
 
     fn schedule_task_by_id(&mut self, id: usize) -> bool {
-        if self.next >= self.descriptor.range.end {
-            false
-        } else {
-            let next = std::cmp::min(
-                self.next + self.descriptor.segment_size_blks,
-                self.descriptor.range.end,
-            );
-            self.task_pool.schedule_segment_rebuild(
-                id,
-                self.next,
-                self.descriptor.clone(),
-            );
-            self.task_pool.active += 1;
-            self.next = next;
-            true
-        }
+        self.copier
+            .next()
+            .map(|blk| {
+                self.task_pool.schedule_segment_rebuild(
+                    id,
+                    blk,
+                    self.copier.copier(),
+                );
+                self.task_pool.active += 1;
+                true
+            })
+            .unwrap_or_default()
     }
 
     async fn await_one_task(&mut self) -> Option<TaskResult> {
@@ -110,7 +113,7 @@ impl RebuildBackend for BdevRebuildJobBackend {
 impl std::fmt::Debug for BdevRebuildJobBackend {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         f.debug_struct("BdevRebuildJob")
-            .field("next", &self.next)
+            .field("next", &self.copier.peek_next())
             .finish()
     }
 }
@@ -130,15 +133,10 @@ impl BdevRebuildJobBackend {
         notify_fn: fn(&str, &str) -> (),
         descriptor: RebuildDescriptor,
     ) -> Result<Self, RebuildError> {
-        let be = Self {
-            next: descriptor.range.start,
+        Ok(Self {
             task_pool,
-            descriptor: Rc::new(descriptor),
+            copier: FullRebuild::new(descriptor),
             notify_fn,
-        };
-
-        info!("{be}: backend created");
-
-        Ok(be)
+        })
     }
 }

io-engine/src/rebuild/mod.rs

@@ -9,9 +9,10 @@ mod rebuild_map;
 mod rebuild_state;
 mod rebuild_stats;
 mod rebuild_task;
+mod rebuilders;
 
 pub use bdev_rebuild::BdevRebuildJob;
-pub use nexus_rebuild::NexusRebuildJob;
+pub use nexus_rebuild::{NexusRebuildJob, NexusRebuildJobStarter};
 use rebuild_descriptor::RebuildDescriptor;
 pub(crate) use rebuild_error::RebuildError;
 use rebuild_job::RebuildOperation;