You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@celeborn.apache.org by et...@apache.org on 2023/05/19 06:05:58 UTC

[incubator-celeborn] branch main updated: [CELEBORN-594] Eliminate Ratis noisy logs. (#1499)

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

ethanfeng 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 ac78afdc4 [CELEBORN-594] Eliminate Ratis noisy logs. (#1499)
ac78afdc4 is described below

commit ac78afdc4e9f89bbbec28a58b5f8c7924c7e00b3
Author: Ethan Feng <et...@apache.org>
AuthorDate: Fri May 19 14:05:52 2023 +0800

    [CELEBORN-594] Eliminate Ratis noisy logs. (#1499)
---
 .../org/apache/celeborn/common/CelebornConf.scala      | 18 ++++++++++++++++++
 conf/log4j2.xml.template                               |  4 ++++
 .../deploy/master/clustermeta/ha/HARaftServer.java     |  7 +++++++
 3 files changed, 29 insertions(+)

diff --git a/common/src/main/scala/org/apache/celeborn/common/CelebornConf.scala b/common/src/main/scala/org/apache/celeborn/common/CelebornConf.scala
index 75d7038b6..ae50b109a 100644
--- a/common/src/main/scala/org/apache/celeborn/common/CelebornConf.scala
+++ b/common/src/main/scala/org/apache/celeborn/common/CelebornConf.scala
@@ -622,6 +622,8 @@ class CelebornConf(loadDefaults: Boolean) extends Cloneable with Logging with Se
   def haMasterRatisRetryCacheExpiryTime: Long = get(HA_MASTER_RATIS_SERVER_RETRY_CACHE_EXPIRY_TIME)
   def haMasterRatisRpcTimeoutMin: Long = get(HA_MASTER_RATIS_RPC_TIMEOUT_MIN)
   def haMasterRatisRpcTimeoutMax: Long = get(HA_MASTER_RATIS_RPC_TIMEOUT_MAX)
+  def haMasterRatisFirstElectionTimeoutMin: Long = get(HA_MASTER_RATIS_FIRSTELECTION_TIMEOUT_MIN)
+  def haMasterRatisFristElectionTimeoutMax: Long = get(HA_MASTER_RATIS_FIRSTELECTION_TIMEOUT_MAX)
   def haMasterRatisNotificationNoLeaderTimeout: Long =
     get(HA_MASTER_RATIS_NOTIFICATION_NO_LEADER_TIMEOUT)
   def haMasterRatisRpcSlownessTimeout: Long = get(HA_MASTER_RATIS_RPC_SLOWNESS_TIMEOUT)
@@ -1735,6 +1737,22 @@ object CelebornConf extends Logging {
       .timeConf(TimeUnit.SECONDS)
       .createWithDefaultString("5s")
 
+  val HA_MASTER_RATIS_FIRSTELECTION_TIMEOUT_MIN: ConfigEntry[Long] =
+    buildConf("celeborn.ha.master.ratis.first.election.timeout.min")
+      .internal
+      .categories("master")
+      .version("0.3.0")
+      .timeConf(TimeUnit.SECONDS)
+      .createWithDefaultString("3s")
+
+  val HA_MASTER_RATIS_FIRSTELECTION_TIMEOUT_MAX: ConfigEntry[Long] =
+    buildConf("celeborn.ha.master.ratis.first.election.timeout.max")
+      .internal
+      .categories("master")
+      .version("0.3.0")
+      .timeConf(TimeUnit.SECONDS)
+      .createWithDefaultString("5s")
+
   val HA_MASTER_RATIS_NOTIFICATION_NO_LEADER_TIMEOUT: ConfigEntry[Long] =
     buildConf("celeborn.ha.master.ratis.raft.server.notification.no-leader.timeout")
       .internal
diff --git a/conf/log4j2.xml.template b/conf/log4j2.xml.template
index b2fdf4ef0..a7dee77c8 100644
--- a/conf/log4j2.xml.template
+++ b/conf/log4j2.xml.template
@@ -53,5 +53,9 @@
             <Appender-ref ref="stdout" level="WARN" />
             <Appender-ref ref="file" level="WARN"/>
         </Logger>
+        <Logger name="org.apache.ratis.server.RaftServerConfigKeys" level="WARN" additivity="false">
+             <Appender-ref ref="stdout" level="WARN" />
+             <Appender-ref ref="file" level="WARN"/>
+        </Logger>
     </Loggers>
 </Configuration>
diff --git a/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/HARaftServer.java b/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/HARaftServer.java
index af9d549af..612a09a43 100644
--- a/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/HARaftServer.java
+++ b/master/src/main/java/org/apache/celeborn/service/deploy/master/clustermeta/ha/HARaftServer.java
@@ -347,6 +347,13 @@ public class HARaftServer {
     RaftServerConfigKeys.Rpc.setTimeoutMin(properties, rpcTimeoutMin);
     RaftServerConfigKeys.Rpc.setTimeoutMax(properties, rpcTimeoutMax);
 
+    TimeDuration firstElectionTimeoutMin =
+        TimeDuration.valueOf(conf.haMasterRatisFirstElectionTimeoutMin(), TimeUnit.SECONDS);
+    TimeDuration firstElectionTimeoutMax =
+        TimeDuration.valueOf(conf.haMasterRatisFristElectionTimeoutMax(), TimeUnit.SECONDS);
+    RaftServerConfigKeys.Rpc.setFirstElectionTimeoutMin(properties, firstElectionTimeoutMin);
+    RaftServerConfigKeys.Rpc.setFirstElectionTimeoutMax(properties, firstElectionTimeoutMax);
+
     // Set the number of maximum cached segments
     RaftServerConfigKeys.Log.setSegmentCacheNumMax(properties, 2);