[incubator-celeborn] branch main updated: [CELEBORN-133] Improve snapshot loading (#1078)

[angerszhuuuu]

This is an automated email from the ASF dual-hosted git repository.

angerszhuuuu pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-celeborn.git


The following commit(s) were added to refs/heads/main by this push:
     new ce86e11d [CELEBORN-133] Improve snapshot loading (#1078)
ce86e11d is described below

commit ce86e11d50fc766da34d690d97b8f09509e91e36
Author: William Song <48054931+szywill...@users.noreply.github.com>
AuthorDate: Wed Dec 14 14:26:31 2022 +0800

    [CELEBORN-133] Improve snapshot loading (#1078)
    
    Co-authored-by: Cheng Pan <pan3...@gmail.com>
---
 .../deploy/master/clustermeta/ha/StateMachine.java  | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git 
a/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/StateMachine.java
 
b/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/StateMachine.java
index d6a82db6..51e8dcba 100644
--- 
a/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/StateMachine.java
+++ 
b/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/StateMachine.java
@@ -17,7 +17,10 @@
 
 package org.apache.celeborn.service.deploy.master.clustermeta.ha;
 
-import static org.apache.ratis.util.LifeCycle.State.*;
+import static org.apache.ratis.util.LifeCycle.State.PAUSED;
+import static org.apache.ratis.util.LifeCycle.State.PAUSING;
+import static org.apache.ratis.util.LifeCycle.State.RUNNING;
+import static org.apache.ratis.util.LifeCycle.State.STARTING;
 
 import java.io.File;
 import java.io.FileNotFoundException;
@@ -207,6 +210,10 @@ public class StateMachine extends BaseStateMachine {
     if (snapshot == null) {
       return;
     }
+    if (snapshot.getTermIndex().compareTo(getLastAppliedTermIndex()) <= 0) {
+      LOG.info("obsolete snapshot provided: {}", snapshot.getTermIndex());
+      return;
+    }
     LOG.info("Loading Snapshot {}.", snapshot);
     final File snapshotFile = snapshot.getFile().getPath().toFile();
     if (!snapshotFile.exists()) {
@@ -216,16 +223,18 @@ public class StateMachine extends BaseStateMachine {
     try {
       setLastAppliedTermIndex(snapshot.getTermIndex());
       install(snapshotFile);
-    } catch (Exception e) {
-      throw new IOException(String.format("Failed to load snapshot %s", 
snapshot), e);
+    } catch (IOException rethrow) {
+      LOG.error("Failed to load snapshot {}", snapshot);
+      throw rethrow;
     }
   }
 
-  private void install(File snapshotFile) {
+  private void install(File snapshotFile) throws IOException {
     try {
       metaHandler.loadSnapShot(snapshotFile);
-    } catch (Exception e) {
-      LOG.warn("Failed to install snapshot!", e);
+    } catch (IOException rethrow) {
+      LOG.warn("Failed to install snapshot!", rethrow);
+      throw rethrow;
     }
     LOG.info("Successfully installed snapshot!");
   }