HDDS-15765. Fail fast when --node-id is omitted for OM compact/defrag on an HA cluster

hadoop-hdds/docs/content/tools/Admin.md

@@ -179,7 +179,7 @@ Note in JSON output mode, field `contToken` won't show up at all in the result i
 The snapshot defrag command triggers the Snapshot Defragmentation Service to run immediately on a specific Ozone Manager node.
 This command manually initiates the snapshot defragmentation process which compacts snapshot data and removes fragmentation to improve storage efficiency.
 
-This command only works on Ozone Manager HA clusters.
+This command only works on Ozone Manager HA clusters. Specify `--node-id` to select which OM to defragment.
 
 ```bash
 $ ozone admin om snapshot defrag --help

hadoop-hdds/docs/content/tools/Repair.md

@@ -182,7 +182,8 @@ CLI to get the status of last trigger quota repair if available.
 
 ### compact
 Compact a column family in the OM DB to clean up tombstones. The compaction happens asynchronously. Requires admin privileges.
-OM should be running for this tool.
+OM should be running for this tool. On an HA OM cluster, specify `--node-id` to select which OM's db to compact.
+
 ```bash
 Usage: ozone repair om compact [-hV] [--dry-run] [--force] [--verbose]
                                --cf=<columnFamilyName> [--node-id=<nodeId>]

hadoop-ozone/cli-admin/src/main/java/org/apache/hadoop/ozone/admin/om/snapshot/DefragSubCommand.java

@@ -22,6 +22,7 @@
 import org.apache.hadoop.hdds.cli.AbstractSubcommand;
 import org.apache.hadoop.hdds.cli.HddsVersionProvider;
 import org.apache.hadoop.hdds.conf.OzoneConfiguration;
+import org.apache.hadoop.ozone.OmUtils;
 import org.apache.hadoop.ozone.admin.om.OmAddressOptions;
 import org.apache.hadoop.ozone.om.helpers.OMNodeDetails;
 import org.apache.hadoop.ozone.om.protocolPB.OMAdminProtocolClientSideImpl;
@@ -64,6 +65,13 @@ public class DefragSubCommand extends AbstractSubcommand implements Callable<Voi
   @Override
   public Void call() throws Exception {
     OzoneConfiguration conf = getOzoneConf();
+
+    if (nodeId == null && OmUtils.isServiceIdsDefined(conf)) {
+      System.err.println("Error: This is an HA OM cluster; specify --node-id "
+          + "to select which OM to defragment.");
+      return null;
+    }
+
     OMNodeDetails omNodeDetails = OMNodeDetails.getOMNodeDetailsFromConf(
         conf, omServiceOption.getServiceID(), nodeId);
 

hadoop-ozone/cli-admin/src/test/java/org/apache/hadoop/ozone/admin/om/snapshot/TestDefragSubCommand.java

@@ -17,10 +17,13 @@
 
 package org.apache.hadoop.ozone.admin.om.snapshot;
 
+import static org.apache.hadoop.ozone.om.OMConfigKeys.OZONE_OM_SERVICE_IDS_KEY;
 import static org.junit.jupiter.api.Assertions.assertTrue;
+import static org.mockito.ArgumentMatchers.anyBoolean;
 import static org.mockito.ArgumentMatchers.eq;
 import static org.mockito.Mockito.doNothing;
 import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.never;
 import static org.mockito.Mockito.verify;
 import static org.mockito.Mockito.when;
 
@@ -54,11 +57,17 @@ public class TestDefragSubCommand {
    */
   private static class TestableDefragSubCommand extends DefragSubCommand {
     private final OMAdminProtocolClientSideImpl mockClient;
+    private final OzoneConfiguration testConf = new OzoneConfiguration();
 
     TestableDefragSubCommand(OMAdminProtocolClientSideImpl mockClient) {
       this.mockClient = mockClient;
     }
 
+    @Override
+    protected OzoneConfiguration getOzoneConf() {
+      return testConf;
+    }
+
     @Override
     protected OMAdminProtocolClientSideImpl createClient(
         OzoneConfiguration conf, OMNodeDetails omNodeDetails) {
@@ -141,6 +150,19 @@ public void testTriggerSnapshotDefragWithServiceIdAndNodeId() throws Exception {
     assertTrue(output.contains("Snapshot defragmentation completed successfully"));
   }
 
+  @Test
+  public void testDefragHAWithoutNodeIdFailsFast() throws Exception {
+    cmd.testConf.set(OZONE_OM_SERVICE_IDS_KEY, "omservice");
+
+    CommandLine c = new CommandLine(cmd);
+    c.parseArgs();
+    cmd.call();
+
+    verify(omAdminClient, never()).triggerSnapshotDefrag(anyBoolean());
+    String error = errContent.toString(DEFAULT_ENCODING);
+    assertTrue(error.contains("specify --node-id"));
+  }
+
   @Test
   public void testTriggerSnapshotDefragWithAllOptions() throws Exception {
     // Test with service-id, node-id, and no-wait options

hadoop-ozone/cli-repair/src/main/java/org/apache/hadoop/ozone/repair/om/CompactOMDB.java

@@ -21,6 +21,7 @@
 import org.apache.hadoop.hdds.cli.HddsVersionProvider;
 import org.apache.hadoop.hdds.conf.OzoneConfiguration;
 import org.apache.hadoop.hdds.utils.db.managed.ManagedCompactRangeOptions;
+import org.apache.hadoop.ozone.OmUtils;
 import org.apache.hadoop.ozone.om.helpers.OMNodeDetails;
 import org.apache.hadoop.ozone.om.protocolPB.OMAdminProtocolClientSideImpl;
 import org.apache.hadoop.ozone.om.service.CompactDBUtil;
@@ -71,6 +72,13 @@ public class CompactOMDB extends RepairTool {
   public void execute() throws Exception {
 
     OzoneConfiguration conf = getOzoneConf();
+
+    if (nodeId == null && OmUtils.isServiceIdsDefined(conf)) {
+      error("This is an HA OM cluster; specify --node-id to select which OM's"
+          + " db to compact.");
+      return;
+    }
+
     OMNodeDetails omNodeDetails = OMNodeDetails.getOMNodeDetailsFromConf(
         conf, omServiceId, nodeId);
 

hadoop-ozone/cli-repair/src/test/java/org/apache/hadoop/ozone/repair/om/TestCompactOMDB.java

@@ -17,12 +17,14 @@
 
 package org.apache.hadoop.ozone.repair.om;
 
+import static org.apache.hadoop.ozone.om.OMConfigKeys.OZONE_OM_SERVICE_IDS_KEY;
 import static org.assertj.core.api.Assertions.assertThat;
 import static org.mockito.ArgumentMatchers.any;
 import static org.mockito.ArgumentMatchers.anyInt;
 import static org.mockito.ArgumentMatchers.eq;
 import static org.mockito.Mockito.mock;
 import static org.mockito.Mockito.mockStatic;
+import static org.mockito.Mockito.never;
 import static org.mockito.Mockito.verify;
 import static org.mockito.Mockito.when;
 
@@ -117,4 +119,14 @@ public void testCompactWhenOMNodeDetailsNotFound() {
 
     assertThat(err.getOutput()).contains("Couldn't determine OM node");
   }
+
+  @Test
+  public void testCompactHAWithoutNodeIdFailsFast() throws Exception {
+    CommandLine cli = new OzoneRepair().getCmd();
+    cli.execute("-D", OZONE_OM_SERVICE_IDS_KEY + "=omservice",
+        "om", "compact", "--column-family", COLUMN_FAMILY);
+
+    verify(omAdminClient, never()).compactOMDB(any(), anyInt());
+    assertThat(err.getOutput()).contains("specify --node-id");
+  }
 }