From 4efcae212bbcbf65027f6fb271cd5f2dd8874c04 Mon Sep 17 00:00:00 2001
From: Radovan Zvoncek <zvo@spotify.com>
Date: Mon, 13 Apr 2015 15:52:23 +0200
Subject: [PATCH] Catch AssertionError if a keyspace doesn't exist

This can happen if somebody changes Cassandra schema if a cluster is
already registered in the Reaper. Without catching, the thread was
silently dying.
---
 .../spotify/reaper/cassandra/JmxProxy.java    | 10 ++++++++--
 .../spotify/reaper/service/RepairRunner.java  | 19 ++++++-------------
 2 files changed, 14 insertions(+), 15 deletions(-)

diff --git a/src/main/java/com/spotify/reaper/cassandra/JmxProxy.java b/src/main/java/com/spotify/reaper/cassandra/JmxProxy.java
index 383a50277..cd51e4e81 100644
--- a/src/main/java/com/spotify/reaper/cassandra/JmxProxy.java
+++ b/src/main/java/com/spotify/reaper/cassandra/JmxProxy.java
@@ -170,9 +170,15 @@ public BigInteger apply(String s) {
         });
   }
 
-  public Map<List<String>, List<String>> getRangeToEndpointMap(String keyspace) {
+  public Map<List<String>, List<String>> getRangeToEndpointMap(String keyspace)
+      throws ReaperException {
     checkNotNull(ssProxy, "Looks like the proxy is not connected");
-    return ssProxy.getRangeToEndpointMap(keyspace);
+    try {
+      return ssProxy.getRangeToEndpointMap(keyspace);
+    } catch (AssertionError e) {
+      LOG.error(e.getMessage());
+      throw new ReaperException(e.getMessage());
+    }
   }
 
   /**
diff --git a/src/main/java/com/spotify/reaper/service/RepairRunner.java b/src/main/java/com/spotify/reaper/service/RepairRunner.java
index ff678a734..c58b88d25 100644
--- a/src/main/java/com/spotify/reaper/service/RepairRunner.java
+++ b/src/main/java/com/spotify/reaper/service/RepairRunner.java
@@ -16,7 +16,6 @@
 import com.google.common.annotations.VisibleForTesting;
 import com.google.common.base.Function;
 import com.google.common.base.Optional;
-
 import com.google.common.collect.Collections2;
 import com.google.common.collect.Lists;
 import com.google.common.util.concurrent.FutureCallback;
@@ -29,7 +28,6 @@
 import com.spotify.reaper.core.RepairRun;
 import com.spotify.reaper.core.RepairSegment;
 import com.spotify.reaper.core.RepairUnit;
-
 import org.joda.time.DateTime;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -38,11 +36,8 @@
 import java.util.Collections;
 import java.util.List;
 import java.util.Map;
-import java.util.concurrent.Future;
 import java.util.concurrent.atomic.AtomicLongArray;
 
-import javax.annotation.Nullable;
-
 public class RepairRunner implements Runnable {
 
   private static final Logger LOG = LoggerFactory.getLogger(RepairRunner.class);
@@ -51,7 +46,6 @@ public class RepairRunner implements Runnable {
   private final long repairRunId;
   private final String clusterName;
   private JmxProxy jmxConnection;
-//  private Long currentlyRunningSegmentId;
   private final AtomicLongArray currentlyRunningSegments;
   private final List<RingRange> parallelRanges;
 
@@ -70,9 +64,8 @@ public RepairRunner(AppContext context, long repairRunId)
     this.clusterName = cluster.get().getName();
     JmxProxy jmx = this.context.jmxConnectionFactory.connectAny(cluster.get());
 
-    int parallelRepairs =
-        getPossibleParallelRepairsCount(
-            jmx.getRangeToEndpointMap(repairUnit.get().getKeyspaceName()));
+    String keyspace = repairUnit.get().getKeyspaceName();
+    int parallelRepairs = getPossibleParallelRepairsCount(jmx.getRangeToEndpointMap(keyspace));
     currentlyRunningSegments = new AtomicLongArray(parallelRepairs);
     for(int i=0;i<parallelRepairs;i++) {
       currentlyRunningSegments.set(i, -1);
@@ -205,10 +198,10 @@ private void end() {
     LOG.info("Repairs for repair run #{} done", repairRunId);
     RepairRun repairRun = context.storage.getRepairRun(repairRunId).get();
     boolean success = context.storage.updateRepairRun(repairRun.with()
-                                                          .runState(RepairRun.RunState.DONE)
-                                                          .endTime(DateTime.now())
-                                                          .lastEvent("All done")
-                                                          .build(repairRun.getId()));
+        .runState(RepairRun.RunState.DONE)
+        .endTime(DateTime.now())
+        .lastEvent("All done")
+        .build(repairRun.getId()));
     if (!success) {
       LOG.error("failed updating repair run " + repairRun.getId());
     }