Implement raft snapshot mechanism by rocksdb checkpoint #1247

Linary · 2020-11-09T10:03:43Z

Change-Id: I761c01da871aac6fe6e906f6150c66c7f5b8f67b

javeme · 2020-11-09T11:17:32Z

...raph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/RaftBackendStoreProvider.java

@@ -129,6 +129,7 @@ public void open(String name) {
    @Override
    public void waitStoreStarted() {
        this.context.initRaftNode();
+        LOG.info("The raft node was inited");


The raft node is initialized

javeme · 2020-11-09T11:28:53Z

hugegraph-rocksdb/src/main/java/com/baidu/hugegraph/backend/store/rocksdb/RocksDBStore.java

+            }
+            LOG.info("The store {} load snapshot succeed", this.store);
+        } catch (RocksDBException e) {
+            throw new BackendException("Failed to reload rocksdb: " + e.toString());


javeme · 2020-11-09T11:30:52Z

hugegraph-rocksdb/src/main/java/com/baidu/hugegraph/backend/store/rocksdb/RocksDBStore.java

@@ -586,21 +581,22 @@ protected Session session(HugeType tableType) {

    @Override
    public void writeSnapshot(String parentPath) {
-        Lock readLock = this.truncateLock.readLock();
-        readLock.lock();
+        Lock writeLock = this.truncateLock.writeLock();


rename to storeLock

javeme · 2020-11-09T11:32:37Z

hugegraph-rocksdb/src/main/java/com/baidu/hugegraph/backend/store/rocksdb/RocksDBStore.java

+             * it may produce dirty data.
+             */
+            for (RocksDBSessions sessions : this.dbs.values()) {
+                if (sessions instanceof RocksDBStdSessions) {


move forceCloseRocksDB to RocksDBSessions

javeme · 2020-11-09T11:33:17Z

hugegraph-rocksdb/src/main/java/com/baidu/hugegraph/backend/store/rocksdb/RocksDBStore.java

+            for (RocksDBSessions sessions : this.dbs.values()) {
+                sessions.reload();
+            }
+            LOG.info("The store {} load snapshot succeed", this.store);


successfully

javeme · 2020-11-09T11:37:31Z

hugegraph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/StoreSnapshotFile.java

                    executor.execute(() -> compressSnapshot(writer, metaBuilder,
                                                            done));
                } else {
                    LOG.error("Failed to save snapshot, path={}, files={}, {}.",
-                              writerPath, writer.listFiles(), throwable);
+                              writerPath, writer.listFiles(), t);


"Failed to save snapshot, path={}, files={}", the exception stack will be printed

javeme · 2020-11-09T11:38:39Z

hugegraph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/StoreSnapshotFile.java

        return CompletableFuture.completedFuture(LocalFileMeta.newBuilder());
    }

-    public void doSnapshotLoad(BackendStore store, String snapshotPath) {
-        store.readSnapshot(snapshotPath);
+    public void doSnapshotLoad(String snapshotPath) {


set private

javeme · 2020-11-09T11:38:57Z

hugegraph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/RaftSharedContext.java

@@ -186,6 +186,10 @@ public StoreType storeType(String store) {
        }
    }

+    public RaftBackendStore[] stores() {


set protected

javeme · 2020-11-09T13:06:03Z

hugegraph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/StoreSnapshotFile.java

@@ -64,7 +64,8 @@ public void save(SnapshotWriter writer, Closure done,
                    executor.execute(() -> compressSnapshot(writer, metaBuilder,
                                                            done));
                } else {
-                    LOG.error("Failed to save snapshot, path={}, files={}, {}.",
+                    LOG.error("Failed to save snapshot, path={}, files={}, " +
+                              "the exception stack will be printed.",


"the exception stack will be printed" is just comment

houzhizhen · 2020-11-10T11:33:24Z

...raph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/RaftBackendStoreProvider.java

@@ -129,8 +129,9 @@ public void open(String name) {
    @Override
    public void waitStoreStarted() {
        this.context.initRaftNode();
+        LOG.info("The raft node is initialized");


remove "is"

prefer keep it

houzhizhen · 2020-11-10T11:33:31Z

...raph-core/src/main/java/com/baidu/hugegraph/backend/store/raft/RaftBackendStoreProvider.java

        this.context.waitRaftNodeStarted();
-        LOG.info("The raft store was started");
+        LOG.info("The raft store is started");


remove "is"

Change-Id: I761c01da871aac6fe6e906f6150c66c7f5b8f67b

Change-Id: Ia423824791a1c6bd611e981facaf7a26c94fc5ca

Change-Id: Ib2c72eab8b91abbba55cca3b309ed91e35654083

Change-Id: I27be0ad4bf78694e98855d77450a97f447832114

Change-Id: I157a1c3b6f893083f507b99a22e717e2b89eb2ce

Change-Id: I205329c07fc8bac0f9731e77ddc842f1366d21ec

Change-Id: I37779caec0fdd50276f8597fa4540df39ad40942

javeme · 2020-11-17T12:52:10Z

...raph-rocksdb/src/main/java/com/baidu/hugegraph/backend/store/rocksdb/RocksDBStdSessions.java

+    public void reload() throws RocksDBException {
+        if (this.rocksdb.isOwningHandle()) {
+            this.rocksdb.close();
+        }


move line 259 to line 256

javeme · 2020-11-17T13:04:05Z

hugegraph-rocksdb/src/main/java/com/baidu/hugegraph/backend/store/rocksdb/RocksDBStore.java

@@ -265,7 +260,7 @@ protected RocksDBSessions open(HugeConfig config, String dataPath,
            sessions = this.openSessionPool(config, dataPath,
                                            walPath, tableNames);
        } catch (RocksDBException e) {
-            RocksDBSessions origin = dbs.get(dataPath);
+            RocksDBSessions origin = this.dbs.get(dataPath);
            if (origin != null) {
                if (e.getMessage().contains("No locks available")) {
                    // Open twice, but we should support keyspace


update comments:
Open twice, copy a RocksDBSessions reference, since from v0.11.2 release we don't support multi graphs share rocksdb instance (before v0.11.2 graphs with different CF-prefix share one rocksdb instance and data path), so each graph has its independent data paths, but multi CFs may share same optimized disk(or optimized disk path).

Change-Id: I5faf30bc530a6ad21fee81fef22614d4e46123a9

codecov · 2020-11-18T07:25:37Z

Codecov Report

Merging #1247 (a6aac8f) into master (b4c767b) will decrease coverage by 6.80%.
The diff coverage is 42.88%.

@@             Coverage Diff              @@
##             master    #1247      +/-   ##
============================================
- Coverage     69.19%   62.38%   -6.81%     
- Complexity     5401     5766     +365     
============================================
  Files           328      374      +46     
  Lines         26334    31329    +4995     
  Branches       3750     4398     +648     
============================================
+ Hits          18222    19545    +1323     
- Misses         6333     9800    +3467     
- Partials       1779     1984     +205

Impacted Files	Coverage Δ	Complexity Δ
...api/src/main/java/com/baidu/hugegraph/api/API.java	`69.04% <ø> (+3.49%)`	`0.00 <0.00> (ø)`
...n/java/com/baidu/hugegraph/api/auth/AccessAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...n/java/com/baidu/hugegraph/api/auth/BelongAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...in/java/com/baidu/hugegraph/api/auth/GroupAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...n/java/com/baidu/hugegraph/api/auth/TargetAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...ain/java/com/baidu/hugegraph/api/auth/UserAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...com/baidu/hugegraph/api/gremlin/GremlinClient.java	`90.90% <ø> (ø)`	`0.00 <0.00> (ø)`
.../java/com/baidu/hugegraph/api/job/ComputerAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...ain/java/com/baidu/hugegraph/api/raft/RaftAPI.java	`0.00% <0.00%> (ø)`	`0.00 <0.00> (?)`
...a/com/baidu/hugegraph/api/schema/EdgeLabelAPI.java	`59.72% <ø> (+2.77%)`	`0.00 <0.00> (ø)`
... and 316 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5a68f50...a6aac8f. Read the comment docs.

* adapt with dbs as member field * Adapt MultiGraphTests Change-Id: I5faf30bc530a6ad21fee81fef22614d4e46123a9

Linary force-pushed the fix-snapshot branch 2 times, most recently from f027175 to 8405036 Compare November 9, 2020 10:13

javeme requested changes Nov 9, 2020

View reviewed changes

javeme reviewed Nov 9, 2020

View reviewed changes

javeme previously approved these changes Nov 10, 2020

View reviewed changes

houzhizhen reviewed Nov 10, 2020

View reviewed changes

Linary dismissed javeme’s stale review via c49a48d November 16, 2020 07:07

Linary force-pushed the fix-snapshot branch from b11a219 to c49a48d Compare November 16, 2020 07:07

houzhizhen previously approved these changes Nov 17, 2020

View reviewed changes

Linary added 6 commits November 17, 2020 15:40

Implement raft snapshot mechanism by rocksdb checkpoint

5254d0c

Change-Id: I761c01da871aac6fe6e906f6150c66c7f5b8f67b

tiny improve

09e4d29

Change-Id: Ia423824791a1c6bd611e981facaf7a26c94fc5ca

tiny improve

10aa243

Change-Id: Ib2c72eab8b91abbba55cca3b309ed91e35654083

tiny improve

e57d3b1

Change-Id: I27be0ad4bf78694e98855d77450a97f447832114

adapt unit test

9c3c49c

Change-Id: I157a1c3b6f893083f507b99a22e717e2b89eb2ce

adapt with dbs as member field

e14b091

Change-Id: I205329c07fc8bac0f9731e77ddc842f1366d21ec

Linary dismissed houzhizhen’s stale review via e14b091 November 17, 2020 07:50

Linary force-pushed the fix-snapshot branch from c49a48d to e14b091 Compare November 17, 2020 07:50

SuppressWarnings

e9de377

Change-Id: I37779caec0fdd50276f8597fa4540df39ad40942

javeme previously approved these changes Nov 17, 2020

View reviewed changes

houzhizhen previously approved these changes Nov 17, 2020

View reviewed changes

zhoney previously approved these changes Nov 17, 2020

View reviewed changes

javeme reviewed Nov 17, 2020

View reviewed changes

Adapt MultiGraphTests

a6aac8f

Change-Id: I5faf30bc530a6ad21fee81fef22614d4e46123a9

Linary dismissed stale reviews from zhoney, houzhizhen, and javeme via a6aac8f November 18, 2020 06:49

javeme approved these changes Nov 18, 2020

View reviewed changes

houzhizhen approved these changes Nov 18, 2020

View reviewed changes

Linary merged commit d3fc084 into master Nov 18, 2020

Linary deleted the fix-snapshot branch November 18, 2020 08:09

tmljob pushed a commit to tmljob/hugegraph that referenced this pull request Dec 10, 2020

Implement raft snapshot mechanism by rocksdb checkpoint (apache#1247)

6c7c47d

* adapt with dbs as member field * Adapt MultiGraphTests Change-Id: I5faf30bc530a6ad21fee81fef22614d4e46123a9

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement raft snapshot mechanism by rocksdb checkpoint #1247

Implement raft snapshot mechanism by rocksdb checkpoint #1247

Linary commented Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

javeme Nov 9, 2020

houzhizhen Nov 10, 2020

javeme Nov 11, 2020

houzhizhen Nov 10, 2020

javeme Nov 17, 2020

javeme Nov 17, 2020

codecov bot commented Nov 18, 2020 •

edited

Implement raft snapshot mechanism by rocksdb checkpoint #1247

Implement raft snapshot mechanism by rocksdb checkpoint #1247

Conversation

Linary commented Nov 9, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov bot commented Nov 18, 2020 • edited

Codecov Report

codecov bot commented Nov 18, 2020 •

edited