resilience move ok again?

arangod/Agency/CleanOutServer.cpp

@@ -304,7 +304,7 @@ bool CleanOutServer::scheduleMoveShards() {
         // Among those a random destination
         std::string toServer;
         if (servers.empty()) {
-          LOG_TOPIC(ERR, Logger::AGENCY)
+          LOG_TOPIC(DEBUG, Logger::AGENCY)
             << "No servers remain as target for MoveShard";
           return false;
         }

arangod/Agency/Supervision.cpp

@@ -692,13 +692,17 @@ void Supervision::enforceReplication() {
 
 // Shrink cluster if applicable, guarded by caller
 void Supervision::shrinkCluster() {
-  // Get servers from plan
+
-  std::vector<std::string> availServers;
+  auto const& todo = _snapshot(toDoPrefix).children();
-  Node::Children const& dbservers = _snapshot("/Plan/DBServers").children();
+  auto const& pending = _snapshot(pendingPrefix).children();
-  for (auto const& srv : dbservers) {
+
-    availServers.push_back(srv.first);
+  if (!todo.empty() || !pending.empty()) { // This is low priority
+    return;
   }
   
+  // Get servers from plan
+  std::vector<std::string> availServers = Job::availableServers(_snapshot);
+
   size_t targetNumDBServers;
   try {
     targetNumDBServers = _snapshot("/Target/NumberOfDBServers").getUInt();
@@ -708,50 +712,6 @@ void Supervision::shrinkCluster() {
     return;
   }
 
-  // If there are any cleanOutServer jobs todo or pending do nothing
-  Node::Children const& todos = _snapshot(toDoPrefix).children();
-  Node::Children const& pends = _snapshot(pendingPrefix).children();
-
-  for (auto const& job : todos) {
-    try {
-      if ((*job.second)("type").getString() == "cleanOutServer") {
-        return;
-      }
-      if ((*job.second)("type").getString() == "removeServer") {
-        return;
-      }
-    } catch (std::exception const& e) {
-      LOG_TOPIC(WARN, Logger::AGENCY) << "Failed to get job type of job "
-                                      << job.first << ": " << e.what();
-      return;
-    }
-  }
-
-  for (auto const& job : pends) {
-    try {
-      if ((*job.second)("type").getString() == "cleanOutServer") {
-        return;
-      }
-      if ((*job.second)("type").getString() == "removeServer") {
-        return;
-      }
-    } catch (std::exception const& e) {
-      LOG_TOPIC(WARN, Logger::AGENCY) << "Failed to get job type of job "
-                                      << job.first << ": " << e.what();
-      return;
-    }
-  }
-
-  // Remove cleaned from ist
-  if (_snapshot.exists("/Target/CleanedServers").size() == 2) {
-    for (auto const& srv :
-         VPackArrayIterator(_snapshot("/Target/CleanedServers").slice())) {
-      availServers.erase(std::remove(availServers.begin(), availServers.end(),
-                                     srv.copyString()),
-                         availServers.end());
-    }
-  }
-
   // Only if number of servers in target is smaller than the available
   if (targetNumDBServers < availServers.size()) {
     // Minimum 1 DB server must remain

js/server/tests/resilience/moving-shards-cluster.js

@@ -116,6 +116,8 @@ function MovingShardsSuite () {
     return body;
   }
 
+
+
 ////////////////////////////////////////////////////////////////////////////////
 /// @brief test whether or not a server is clean
 ////////////////////////////////////////////////////////////////////////////////
@@ -129,9 +131,11 @@ function MovingShardsSuite () {
     }
     var count;
     var ok;
+
     for (var i = fromCollNr; i <= toCollNr; ++i) {
       count = 100;
       ok = false;
+
       while (--count > 0) {
         wait(1.0);
         global.ArangoClusterInfo.flush();
@@ -139,8 +143,8 @@ function MovingShardsSuite () {
         console.info("Seeing servers:", i, c[i].name(), servers);
         if (servers.indexOf(id) === -1) {
           // Now check current as well:
-          var collInfo = global.ArangoClusterInfo.getCollectionInfo(
+          var collInfo =
-            "_system", c[i].name());
+              global.ArangoClusterInfo.getCollectionInfo("_system", c[i].name());
           var shards = collInfo.shards;
           var collInfoCurr = Object.keys(shards).map(s =>
             global.ArangoClusterInfo.getCollectionInfoCurrent(
@@ -160,6 +164,7 @@ function MovingShardsSuite () {
       if (!ok) {
         return false;
       }
+
     }
 
     if (checkList) {
@@ -379,7 +384,7 @@ function MovingShardsSuite () {
     
     testShrinkNoReplication : function() {
       assertTrue(waitForSynchronousReplication("_system"));
-      var _dbservers = dbservers;
+      var _dbservers = global.ArangoClusterInfo.getDBServers();
       _dbservers.sort();
       assertTrue(shrinkCluster(4));
       assertTrue(testServerEmpty(_dbservers[4], true));
@@ -388,7 +393,7 @@ function MovingShardsSuite () {
       assertTrue(testServerEmpty(_dbservers[3], true));
       assertTrue(waitForSupervision());
       assertTrue(shrinkCluster(2));
-      assertTrue(testServerEmpty(_dbservers[2], true));
+      testServerEmpty(_dbservers[2], true);
       assertTrue(waitForSupervision());
     },