[BugFix] Fix the bug in the FE clone scheduling that causes an infini…

…te loop (backport #50561) (#50787) Signed-off-by: sevev <[email protected]> Co-authored-by: zhangqiang <[email protected]>
StarRocks · Sep 9, 2024 · 7207d3a · 7207d3a
1 parent 7209a49
commit 7207d3a
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 1 deletion.
diff --git a/fe/fe-core/src/main/java/com/starrocks/clone/TabletSchedCtx.java b/fe/fe-core/src/main/java/com/starrocks/clone/TabletSchedCtx.java
@@ -637,7 +637,11 @@ public boolean chooseDestReplicaForVersionIncomplete(Map<Long, PathSlot> backend
             } else if (replica.getLastSuccessVersion() > replica.getLastFailedVersion()) {
                 chosenReplica = replica;
                 break;
-            } else if (replica.getLastFailedVersion() < chosenReplica.getLastFailedVersion()) {
+            } else if (replica.getVersion() < chosenReplica.getVersion()) {
+                // we should first repair the lower version replica
+                chosenReplica = replica;
+            } else if (replica.getVersion() == chosenReplica.getVersion() &&
+                       replica.getLastFailedVersion() < chosenReplica.getLastFailedVersion()) {
                 // it's better to select a low last failed version replica
                 chosenReplica = replica;
             }

diff --git a/fe/fe-core/src/test/java/com/starrocks/clone/TabletSchedCtxTest.java b/fe/fe-core/src/test/java/com/starrocks/clone/TabletSchedCtxTest.java
@@ -69,6 +69,7 @@ public class TabletSchedCtxTest {
             .createType(PrimitiveType.TINYINT), true, null, "", "key1"));
 
     private static int TABLET_ID_1 = 50000;
+    private static int TABLET_ID_2 = 51000;
 
     private Backend be1;
     private Backend be2;