Enable supression of dead servers in Monitor by resource group (apach…

…e#5226) Closes apache#5162 Co-authored-by: Dom G. <[email protected]>
dlmarion · Jan 8, 2025 · d2f25d3 · d2f25d3
1 parent 05fe071
commit d2f25d3
Show file tree

Hide file tree

Showing 3 changed files with 28 additions and 2 deletions.
diff --git a/core/src/main/java/org/apache/accumulo/core/conf/Property.java b/core/src/main/java/org/apache/accumulo/core/conf/Property.java
@@ -830,6 +830,12 @@ public enum Property {
           + " The resources that are used by default can be seen in"
           + " `accumulo/server/monitor/src/main/resources/templates/default.ftl`.",
       "2.0.0"),
+  MONITOR_DEAD_LIST_RG_EXCLUSIONS("monitor.dead.server.rg.exclusions", "", PropertyType.STRING,
+      "The Monitor displays information about servers that it believes have died recently."
+          + " This property accepts a comma separated list of resource group names. If"
+          + " the dead servers resource group matches a resource group in this list,"
+          + " then it will be suppressed from the dead servers list in the monitor.",
+      "4.0.0"),
   // per table properties
   TABLE_PREFIX("table.", null, PropertyType.PREFIX,
       "Properties in this category affect tablet server treatment of tablets,"

diff --git a/server/monitor/pom.xml b/server/monitor/pom.xml
@@ -76,6 +76,10 @@
       <groupId>org.apache.accumulo</groupId>
       <artifactId>accumulo-start</artifactId>
     </dependency>
+    <dependency>
+      <groupId>org.apache.commons</groupId>
+      <artifactId>commons-lang3</artifactId>
+    </dependency>
     <dependency>
       <groupId>org.apache.logging.log4j</groupId>
       <artifactId>log4j-web</artifactId>

diff --git a/server/monitor/src/main/java/org/apache/accumulo/monitor/rest/manager/ManagerResource.java b/server/monitor/src/main/java/org/apache/accumulo/monitor/rest/manager/ManagerResource.java
@@ -20,6 +20,7 @@
 
 import java.lang.management.ManagementFactory;
 import java.util.ArrayList;
+import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
 import java.util.Map.Entry;
@@ -32,6 +33,7 @@
 import jakarta.ws.rs.core.MediaType;
 
 import org.apache.accumulo.core.client.admin.servers.ServerId;
+import org.apache.accumulo.core.conf.Property;
 import org.apache.accumulo.core.gc.thrift.GCStatus;
 import org.apache.accumulo.core.manager.thrift.DeadServer;
 import org.apache.accumulo.core.manager.thrift.ManagerMonitorInfo;
@@ -46,6 +48,7 @@
 import org.apache.accumulo.monitor.rest.tservers.ServerShuttingDownInformation;
 import org.apache.accumulo.monitor.rest.tservers.ServersShuttingDown;
 import org.apache.accumulo.server.manager.state.TabletServerState;
+import org.apache.commons.lang3.StringUtils;
 
 /**
  * Responsible for generating a new Manager information JSON object
@@ -173,10 +176,23 @@ public static DeadServerList getDeadTservers(Monitor monitor) {
     }
 
     DeadServerList deadServers = new DeadServerList();
+
+    String prop = monitor.getConfiguration().get(Property.MONITOR_DEAD_LIST_RG_EXCLUSIONS);
+    Set<String> exclusions = null;
+    if (StringUtils.isNotBlank(prop)) {
+      String[] rgs = prop.split(",");
+      exclusions = new HashSet<>(rgs.length);
+      for (String s : rgs) {
+        exclusions.add(s.trim());
+      }
+    }
+
     // Add new dead servers to the list
     for (DeadServer dead : mmi.deadTabletServers) {
-      deadServers
-          .addDeadServer(new DeadServerInformation(dead.server, dead.lastStatus, dead.status));
+      if (exclusions == null || !exclusions.contains(dead.getResourceGroup())) {
+        deadServers
+            .addDeadServer(new DeadServerInformation(dead.server, dead.lastStatus, dead.status));
+      }
     }
     return deadServers;
   }