/**
* Generate the assignment plan for the existing table
*
* @param tableName
* @param assignmentSnapshot
* @param regionLocalityMap
* @param plan
* @param munkresForSecondaryAndTertiary if set on true the assignment plan for the tertiary and
* secondary will be generated with Munkres algorithm, otherwise will be generated using
* placeSecondaryAndTertiaryRS
* @throws IOException
*/
private void genAssignmentPlan(
TableName tableName,
SnapshotOfRegionAssignmentFromMeta assignmentSnapshot,
Map<String, Map<String, Float>> regionLocalityMap,
FavoredNodesPlan plan,
boolean munkresForSecondaryAndTertiary)
throws IOException {
// Get the all the regions for the current table
List<HRegionInfo> regions = assignmentSnapshot.getTableToRegionMap().get(tableName);
int numRegions = regions.size();
// Get the current assignment map
Map<HRegionInfo, ServerName> currentAssignmentMap =
assignmentSnapshot.getRegionToRegionServerMap();
// Get the all the region servers
List<ServerName> servers = new ArrayList<ServerName>();
try (Admin admin = this.connection.getAdmin()) {
servers.addAll(admin.getClusterStatus().getServers());
}
LOG.info(
"Start to generate assignment plan for "
+ numRegions
+ " regions from table "
+ tableName
+ " with "
+ servers.size()
+ " region servers");
int slotsPerServer = (int) Math.ceil((float) numRegions / servers.size());
int regionSlots = slotsPerServer * servers.size();
// Compute the primary, secondary and tertiary costs for each region/server
// pair. These costs are based only on node locality and rack locality, and
// will be modified later.
float[][] primaryCost = new float[numRegions][regionSlots];
float[][] secondaryCost = new float[numRegions][regionSlots];
float[][] tertiaryCost = new float[numRegions][regionSlots];
if (this.enforceLocality && regionLocalityMap != null) {
// Transform the locality mapping into a 2D array, assuming that any
// unspecified locality value is 0.
float[][] localityPerServer = new float[numRegions][regionSlots];
for (int i = 0; i < numRegions; i++) {
Map<String, Float> serverLocalityMap =
regionLocalityMap.get(regions.get(i).getEncodedName());
if (serverLocalityMap == null) {
continue;
}
for (int j = 0; j < servers.size(); j++) {
String serverName = servers.get(j).getHostname();
if (serverName == null) {
continue;
}
Float locality = serverLocalityMap.get(serverName);
if (locality == null) {
continue;
}
for (int k = 0; k < slotsPerServer; k++) {
// If we can't find the locality of a region to a server, which occurs
// because locality is only reported for servers which have some
// blocks of a region local, then the locality for that pair is 0.
localityPerServer[i][j * slotsPerServer + k] = locality.floatValue();
}
}
}
// Compute the total rack locality for each region in each rack. The total
// rack locality is the sum of the localities of a region on all servers in
// a rack.
Map<String, Map<HRegionInfo, Float>> rackRegionLocality =
new HashMap<String, Map<HRegionInfo, Float>>();
for (int i = 0; i < numRegions; i++) {
HRegionInfo region = regions.get(i);
for (int j = 0; j < regionSlots; j += slotsPerServer) {
String rack = rackManager.getRack(servers.get(j / slotsPerServer));
Map<HRegionInfo, Float> rackLocality = rackRegionLocality.get(rack);
if (rackLocality == null) {
rackLocality = new HashMap<HRegionInfo, Float>();
rackRegionLocality.put(rack, rackLocality);
}
Float localityObj = rackLocality.get(region);
float locality = localityObj == null ? 0 : localityObj.floatValue();
locality += localityPerServer[i][j];
rackLocality.put(region, locality);
}
}
for (int i = 0; i < numRegions; i++) {
for (int j = 0; j < regionSlots; j++) {
String rack = rackManager.getRack(servers.get(j / slotsPerServer));
Float totalRackLocalityObj = rackRegionLocality.get(rack).get(regions.get(i));
float totalRackLocality =
totalRackLocalityObj == null ? 0 : totalRackLocalityObj.floatValue();
// Primary cost aims to favor servers with high node locality and low
// rack locality, so that secondaries and tertiaries can be chosen for
// nodes with high rack locality. This might give primaries with
// slightly less locality at first compared to a cost which only
// considers the node locality, but should be better in the long run.
primaryCost[i][j] = 1 - (2 * localityPerServer[i][j] - totalRackLocality);
// Secondary cost aims to favor servers with high node locality and high
// rack locality since the tertiary will be chosen from the same rack as
// the secondary. This could be negative, but that is okay.
secondaryCost[i][j] = 2 - (localityPerServer[i][j] + totalRackLocality);
// Tertiary cost is only concerned with the node locality. It will later
// be restricted to only hosts on the same rack as the secondary.
tertiaryCost[i][j] = 1 - localityPerServer[i][j];
}
}
}
if (this.enforceMinAssignmentMove && currentAssignmentMap != null) {
// We want to minimize the number of regions which move as the result of a
// new assignment. Therefore, slightly penalize any placement which is for
// a host that is not currently serving the region.
for (int i = 0; i < numRegions; i++) {
for (int j = 0; j < servers.size(); j++) {
ServerName currentAddress = currentAssignmentMap.get(regions.get(i));
if (currentAddress != null && !currentAddress.equals(servers.get(j))) {
for (int k = 0; k < slotsPerServer; k++) {
primaryCost[i][j * slotsPerServer + k] += NOT_CURRENT_HOST_PENALTY;
}
}
}
}
}
// Artificially increase cost of last slot of each server to evenly
// distribute the slop, otherwise there will be a few servers with too few
// regions and many servers with the max number of regions.
for (int i = 0; i < numRegions; i++) {
for (int j = 0; j < regionSlots; j += slotsPerServer) {
primaryCost[i][j] += LAST_SLOT_COST_PENALTY;
secondaryCost[i][j] += LAST_SLOT_COST_PENALTY;
tertiaryCost[i][j] += LAST_SLOT_COST_PENALTY;
}
}
RandomizedMatrix randomizedMatrix = new RandomizedMatrix(numRegions, regionSlots);
primaryCost = randomizedMatrix.transform(primaryCost);
int[] primaryAssignment = new MunkresAssignment(primaryCost).solve();
primaryAssignment = randomizedMatrix.invertIndices(primaryAssignment);
// Modify the secondary and tertiary costs for each region/server pair to
// prevent a region from being assigned to the same rack for both primary
// and either one of secondary or tertiary.
for (int i = 0; i < numRegions; i++) {
int slot = primaryAssignment[i];
String rack = rackManager.getRack(servers.get(slot / slotsPerServer));
for (int k = 0; k < servers.size(); k++) {
if (!rackManager.getRack(servers.get(k)).equals(rack)) {
continue;
}
if (k == slot / slotsPerServer) {
// Same node, do not place secondary or tertiary here ever.
for (int m = 0; m < slotsPerServer; m++) {
secondaryCost[i][k * slotsPerServer + m] = MAX_COST;
tertiaryCost[i][k * slotsPerServer + m] = MAX_COST;
}
} else {
// Same rack, do not place secondary or tertiary here if possible.
for (int m = 0; m < slotsPerServer; m++) {
secondaryCost[i][k * slotsPerServer + m] = AVOID_COST;
tertiaryCost[i][k * slotsPerServer + m] = AVOID_COST;
}
}
}
}
if (munkresForSecondaryAndTertiary) {
randomizedMatrix = new RandomizedMatrix(numRegions, regionSlots);
secondaryCost = randomizedMatrix.transform(secondaryCost);
int[] secondaryAssignment = new MunkresAssignment(secondaryCost).solve();
secondaryAssignment = randomizedMatrix.invertIndices(secondaryAssignment);
// Modify the tertiary costs for each region/server pair to ensure that a
// region is assigned to a tertiary server on the same rack as its secondary
// server, but not the same server in that rack.
for (int i = 0; i < numRegions; i++) {
int slot = secondaryAssignment[i];
String rack = rackManager.getRack(servers.get(slot / slotsPerServer));
for (int k = 0; k < servers.size(); k++) {
if (k == slot / slotsPerServer) {
// Same node, do not place tertiary here ever.
for (int m = 0; m < slotsPerServer; m++) {
tertiaryCost[i][k * slotsPerServer + m] = MAX_COST;
}
} else {
if (rackManager.getRack(servers.get(k)).equals(rack)) {
continue;
}
// Different rack, do not place tertiary here if possible.
for (int m = 0; m < slotsPerServer; m++) {
tertiaryCost[i][k * slotsPerServer + m] = AVOID_COST;
}
}
}
}
randomizedMatrix = new RandomizedMatrix(numRegions, regionSlots);
tertiaryCost = randomizedMatrix.transform(tertiaryCost);
int[] tertiaryAssignment = new MunkresAssignment(tertiaryCost).solve();
tertiaryAssignment = randomizedMatrix.invertIndices(tertiaryAssignment);
for (int i = 0; i < numRegions; i++) {
List<ServerName> favoredServers =
new ArrayList<ServerName>(FavoredNodeAssignmentHelper.FAVORED_NODES_NUM);
ServerName s = servers.get(primaryAssignment[i] / slotsPerServer);
favoredServers.add(
ServerName.valueOf(s.getHostname(), s.getPort(), ServerName.NON_STARTCODE));
s = servers.get(secondaryAssignment[i] / slotsPerServer);
favoredServers.add(
ServerName.valueOf(s.getHostname(), s.getPort(), ServerName.NON_STARTCODE));
s = servers.get(tertiaryAssignment[i] / slotsPerServer);
favoredServers.add(
ServerName.valueOf(s.getHostname(), s.getPort(), ServerName.NON_STARTCODE));
// Update the assignment plan
plan.updateAssignmentPlan(regions.get(i), favoredServers);
}
LOG.info(
"Generated the assignment plan for "
+ numRegions
+ " regions from table "
+ tableName
+ " with "
+ servers.size()
+ " region servers");
LOG.info("Assignment plan for secondary and tertiary generated " + "using MunkresAssignment");
} else {
Map<HRegionInfo, ServerName> primaryRSMap = new HashMap<HRegionInfo, ServerName>();
for (int i = 0; i < numRegions; i++) {
primaryRSMap.put(regions.get(i), servers.get(primaryAssignment[i] / slotsPerServer));
}
FavoredNodeAssignmentHelper favoredNodeHelper =
new FavoredNodeAssignmentHelper(servers, conf);
favoredNodeHelper.initialize();
Map<HRegionInfo, ServerName[]> secondaryAndTertiaryMap =
favoredNodeHelper.placeSecondaryAndTertiaryWithRestrictions(primaryRSMap);
for (int i = 0; i < numRegions; i++) {
List<ServerName> favoredServers =
new ArrayList<ServerName>(FavoredNodeAssignmentHelper.FAVORED_NODES_NUM);
HRegionInfo currentRegion = regions.get(i);
ServerName s = primaryRSMap.get(currentRegion);
favoredServers.add(
ServerName.valueOf(s.getHostname(), s.getPort(), ServerName.NON_STARTCODE));
ServerName[] secondaryAndTertiary = secondaryAndTertiaryMap.get(currentRegion);
s = secondaryAndTertiary[0];
favoredServers.add(
ServerName.valueOf(s.getHostname(), s.getPort(), ServerName.NON_STARTCODE));
s = secondaryAndTertiary[1];
favoredServers.add(
ServerName.valueOf(s.getHostname(), s.getPort(), ServerName.NON_STARTCODE));
// Update the assignment plan
plan.updateAssignmentPlan(regions.get(i), favoredServers);
}
LOG.info(
"Generated the assignment plan for "
+ numRegions
+ " regions from table "
+ tableName
+ " with "
+ servers.size()
+ " region servers");
LOG.info(
"Assignment plan for secondary and tertiary generated "
+ "using placeSecondaryAndTertiaryWithRestrictions method");
}
}
public void printStats() throws IOException {
Admin admin = connection.getAdmin();
ClusterStatus status =
admin.getClusterStatus(); // co ClusterStatusExample-1-GetStatus Get the cluster status.
System.out.println("Cluster Status:\n--------------");
System.out.println("HBase Version: " + status.getHBaseVersion());
System.out.println("Version: " + status.getVersion());
System.out.println("Cluster ID: " + status.getClusterId());
System.out.println("Master: " + status.getMaster());
System.out.println("No. Backup Masters: " + status.getBackupMastersSize());
System.out.println("Backup Masters: " + status.getBackupMasters());
System.out.println("No. Live Servers: " + status.getServersSize());
System.out.println("Servers: " + status.getServers());
System.out.println("No. Dead Servers: " + status.getDeadServers());
System.out.println("Dead Servers: " + status.getDeadServerNames());
System.out.println("No. Regions: " + status.getRegionsCount());
System.out.println("Regions in Transition: " + status.getRegionsInTransition());
System.out.println("No. Requests: " + status.getRequestsCount());
System.out.println("Avg Load: " + status.getAverageLoad());
System.out.println("Balancer On: " + status.getBalancerOn());
System.out.println("Is Balancer On: " + status.isBalancerOn());
System.out.println("Master Coprocessors: " + Arrays.asList(status.getMasterCoprocessors()));
System.out.println("\nServer Info:\n--------------");
for (ServerName server :
status
.getServers()) { // co ClusterStatusExample-2-ServerInfo Iterate over the included
// server instances.
System.out.println("Hostname: " + server.getHostname());
System.out.println("Host and Port: " + server.getHostAndPort());
System.out.println("Server Name: " + server.getServerName());
System.out.println("RPC Port: " + server.getPort());
System.out.println("Start Code: " + server.getStartcode());
ServerLoad load =
status.getLoad(
server); // co ClusterStatusExample-3-ServerLoad Retrieve the load details for the
// current server.
System.out.println("\nServer Load:\n--------------");
System.out.println("Info Port: " + load.getInfoServerPort());
System.out.println("Load: " + load.getLoad());
System.out.println("Max Heap (MB): " + load.getMaxHeapMB());
System.out.println("Used Heap (MB): " + load.getUsedHeapMB());
System.out.println("Memstore Size (MB): " + load.getMemstoreSizeInMB());
System.out.println("No. Regions: " + load.getNumberOfRegions());
System.out.println("No. Requests: " + load.getNumberOfRequests());
System.out.println("Total No. Requests: " + load.getTotalNumberOfRequests());
System.out.println("No. Requests per Sec: " + load.getRequestsPerSecond());
System.out.println("No. Read Requests: " + load.getReadRequestsCount());
System.out.println("No. Write Requests: " + load.getWriteRequestsCount());
System.out.println("No. Stores: " + load.getStores());
System.out.println("Store Size Uncompressed (MB): " + load.getStoreUncompressedSizeMB());
System.out.println("No. Storefiles: " + load.getStorefiles());
System.out.println("Storefile Size (MB): " + load.getStorefileSizeInMB());
System.out.println("Storefile Index Size (MB): " + load.getStorefileIndexSizeInMB());
System.out.println("Root Index Size: " + load.getRootIndexSizeKB());
System.out.println("Total Bloom Size: " + load.getTotalStaticBloomSizeKB());
System.out.println("Total Index Size: " + load.getTotalStaticIndexSizeKB());
System.out.println("Current Compacted Cells: " + load.getCurrentCompactedKVs());
System.out.println("Total Compacting Cells: " + load.getTotalCompactingKVs());
System.out.println("Coprocessors1: " + Arrays.asList(load.getRegionServerCoprocessors()));
System.out.println("Coprocessors2: " + Arrays.asList(load.getRsCoprocessors()));
System.out.println("Replication Load Sink: " + load.getReplicationLoadSink());
System.out.println("Replication Load Source: " + load.getReplicationLoadSourceList());
System.out.println("\nRegion Load:\n--------------");
for (Map.Entry<byte[], RegionLoad>
entry : // co ClusterStatusExample-4-Regions Iterate over the region details of the
// current server.
load.getRegionsLoad().entrySet()) {
System.out.println("Region: " + Bytes.toStringBinary(entry.getKey()));
RegionLoad regionLoad =
entry
.getValue(); // co ClusterStatusExample-5-RegionLoad Get the load details for the
// current region.
System.out.println("Name: " + Bytes.toStringBinary(regionLoad.getName()));
System.out.println("Name (as String): " + regionLoad.getNameAsString());
System.out.println("No. Requests: " + regionLoad.getRequestsCount());
System.out.println("No. Read Requests: " + regionLoad.getReadRequestsCount());
System.out.println("No. Write Requests: " + regionLoad.getWriteRequestsCount());
System.out.println("No. Stores: " + regionLoad.getStores());
System.out.println("No. Storefiles: " + regionLoad.getStorefiles());
System.out.println("Data Locality: " + regionLoad.getDataLocality());
System.out.println("Storefile Size (MB): " + regionLoad.getStorefileSizeMB());
System.out.println("Storefile Index Size (MB): " + regionLoad.getStorefileIndexSizeMB());
System.out.println("Memstore Size (MB): " + regionLoad.getMemStoreSizeMB());
System.out.println("Root Index Size: " + regionLoad.getRootIndexSizeKB());
System.out.println("Total Bloom Size: " + regionLoad.getTotalStaticBloomSizeKB());
System.out.println("Total Index Size: " + regionLoad.getTotalStaticIndexSizeKB());
System.out.println("Current Compacted Cells: " + regionLoad.getCurrentCompactedKVs());
System.out.println("Total Compacting Cells: " + regionLoad.getTotalCompactingKVs());
System.out.println();
}
}
}