/**
* start a topology: set active status of the topology
*
* @param topologyName
* @param stormClusterState
* @param topologyId
* @throws Exception
*/
public void setTopologyStatus(TopologyAssignEvent event) throws Exception {
StormClusterState stormClusterState = nimbusData.getStormClusterState();
String topologyId = event.getTopologyId();
String topologyName = event.getTopologyName();
String group = event.getGroup();
StormStatus status = new StormStatus(StatusType.active);
if (event.getOldStatus() != null) {
status = event.getOldStatus();
}
StormBase stormBase = stormClusterState.storm_base(topologyId, null);
if (stormBase == null) {
stormBase = new StormBase(topologyName, TimeUtils.current_time_secs(), status, group);
stormClusterState.activate_storm(topologyId, stormBase);
} else {
stormClusterState.update_storm(topologyId, status);
// here exist one hack operation
// when monitor/rebalance/startup topologyName is null
if (topologyName == null) {
event.setTopologyName(stormBase.getStormName());
}
}
LOG.info("Update " + topologyId + " " + status);
}
// Utility functions, just used here
public static Long tuple_time_delta(RotatingMap<Tuple, Long> start_times, Tuple tuple) {
Long start_time = (Long) start_times.remove(tuple);
if (start_time != null) {
return TimeUtils.time_delta_ms(start_time);
}
return null;
}
public TaskHeartbeatTrigger(
Map conf,
String name,
DisruptorQueue queue,
BlockingQueue<Object> controlQueue,
int taskId,
String componentId,
TopologyContext sysTopologyCtx,
ITaskReportErr reportError) {
this.name = name;
this.queue = queue;
this.controlQueue = controlQueue;
this.opCode = TimerConstants.TASK_HEARTBEAT;
this.taskId = taskId;
this.componentId = componentId;
this.sysTopologyCtx = sysTopologyCtx;
this.frequence = JStormUtils.parseInt(conf.get(Config.TASK_HEARTBEAT_FREQUENCY_SECS), 10);
this.firstTime = frequence;
this.executeThreadHbTime = TimeUtils.current_time_secs();
this.taskHbTimeout = JStormUtils.parseInt(conf.get(Config.NIMBUS_TASK_TIMEOUT_SECS), 180);
this.intervalCheck = new IntervalCheck();
this.intervalCheck.setInterval(taskHbTimeout);
this.intervalCheck.start();
this.reportError = reportError;
this.uptime = new UptimeComputer();
}
/**
* @param existingAssignment
* @param taskWorkerSlot
* @return
* @throws Exception
*/
public static Map<Integer, Integer> getTaskStartTimes(
TopologyAssignContext context,
NimbusData nimbusData,
String topologyId,
Assignment existingAssignment,
Set<ResourceWorkerSlot> workers)
throws Exception {
Map<Integer, Integer> startTimes = new TreeMap<Integer, Integer>();
if (context.getAssignType() == TopologyAssignContext.ASSIGN_TYPE_NEW) {
int nowSecs = TimeUtils.current_time_secs();
for (ResourceWorkerSlot worker : workers) {
for (Integer changedTaskId : worker.getTasks()) {
startTimes.put(changedTaskId, nowSecs);
}
}
return startTimes;
}
Set<ResourceWorkerSlot> oldWorkers = new HashSet<ResourceWorkerSlot>();
if (existingAssignment != null) {
Map<Integer, Integer> taskStartTimeSecs = existingAssignment.getTaskStartTimeSecs();
if (taskStartTimeSecs != null) {
startTimes.putAll(taskStartTimeSecs);
}
if (existingAssignment.getWorkers() != null) {
oldWorkers = existingAssignment.getWorkers();
}
}
StormClusterState zkClusterState = nimbusData.getStormClusterState();
Set<Integer> changeTaskIds = getChangeTaskIds(oldWorkers, workers);
int nowSecs = TimeUtils.current_time_secs();
for (Integer changedTaskId : changeTaskIds) {
startTimes.put(changedTaskId, nowSecs);
zkClusterState.remove_task_heartbeat(topologyId, changedTaskId);
}
LOG.info("Task assignment has been changed " + changeTaskIds);
return startTimes;
}
private void checkExecuteThreadHb() {
long currentTime = TimeUtils.current_time_secs();
if (currentTime - executeThreadHbTime > taskHbTimeout) {
String error =
"No response from Task-" + taskId + ", last report time(sec) is " + executeThreadHbTime;
reportError.report(error);
}
}
@Override
public void onEvent(Object event, long sequence, boolean endOfBatch) throws Exception {
if (event == null) {
return;
}
boltExeTimer.start();
try {
if (event instanceof RotatingMapTrigger.Tick) {
// don't check the timetick name to improve performance
Map<Tuple, Long> timeoutMap = tuple_start_times.rotate();
if (ackerNum > 0) {
// only when acker is enable
for (Entry<Tuple, Long> entry : timeoutMap.entrySet()) {
Tuple input = entry.getKey();
task_stats.bolt_failed_tuple(input.getSourceComponent(), input.getSourceStreamId());
}
}
return;
}
Tuple tuple = (Tuple) event;
task_stats.recv_tuple(tuple.getSourceComponent(), tuple.getSourceStreamId());
tuple_start_times.put(tuple, System.currentTimeMillis());
try {
bolt.execute(tuple);
} catch (Throwable e) {
error = e;
LOG.error("bolt execute error ", e);
report_error.report(e);
}
if (ackerNum == 0) {
// only when acker is disable
// get tuple process latency
Long start_time = (Long) tuple_start_times.remove(tuple);
if (start_time != null) {
Long delta = TimeUtils.time_delta_ms(start_time);
task_stats.bolt_acked_tuple(tuple.getSourceComponent(), tuple.getSourceStreamId(), delta);
}
}
} finally {
boltExeTimer.stop();
}
}
/**
* Get TopologyInfo, it contain all data of the topology running status
*
* @return TopologyInfo
*/
@Override
public TopologyInfo getTopologyInfo(String topologyId) throws NotAliveException, TException {
TopologyInfo topologyInfo = new TopologyInfo();
StormClusterState stormClusterState = data.getStormClusterState();
try {
// get topology's StormBase
StormBase base = stormClusterState.storm_base(topologyId, null);
if (base == null) {
throw new NotAliveException("No topology of " + topologyId);
}
topologyInfo.set_id(topologyId);
topologyInfo.set_name(base.getStormName());
topologyInfo.set_uptime_secs(TimeUtils.time_delta(base.getLanchTimeSecs()));
topologyInfo.set_status(base.getStatusString());
// get topology's Assignment
Assignment assignment = stormClusterState.assignment_info(topologyId, null);
if (assignment == null) {
throw new TException("Failed to get StormBase from ZK of " + topologyId);
}
// get topology's map<taskId, componentId>
Map<Integer, String> taskInfo = Cluster.topology_task_info(stormClusterState, topologyId);
List<TaskSummary> tasks =
NimbusUtils.mkTaskSummary(stormClusterState, assignment, taskInfo, topologyId);
topologyInfo.set_tasks(tasks);
return topologyInfo;
} catch (TException e) {
LOG.info("Failed to get topologyInfo " + topologyId, e);
throw e;
} catch (Exception e) {
LOG.info("Failed to get topologyInfo " + topologyId, e);
throw new TException("Failed to get topologyInfo" + topologyId);
}
}
@Override
public SupervisorWorkers getSupervisorWorkers(String host) throws NotAliveException, TException {
try {
StormClusterState stormClusterState = data.getStormClusterState();
String supervisorId = null;
SupervisorInfo supervisorInfo = null;
String ip = NetWorkUtils.host2Ip(host);
String hostName = NetWorkUtils.ip2Host(host);
// all supervisors
Map<String, SupervisorInfo> supervisorInfos =
Cluster.allSupervisorInfo(stormClusterState, null);
for (Entry<String, SupervisorInfo> entry : supervisorInfos.entrySet()) {
SupervisorInfo info = entry.getValue();
if (info.getHostName().equals(hostName) || info.getHostName().equals(ip)) {
supervisorId = entry.getKey();
supervisorInfo = info;
break;
}
}
if (supervisorId == null) {
throw new TException("No supervisor of " + host);
}
Map<String, Assignment> assignments = new HashMap<String, Assignment>();
// get all active topology's StormBase
Map<String, StormBase> bases = Cluster.topology_bases(stormClusterState);
for (Entry<String, StormBase> entry : bases.entrySet()) {
String topologyId = entry.getKey();
StormBase base = entry.getValue();
Assignment assignment = stormClusterState.assignment_info(topologyId, null);
if (assignment == null) {
LOG.error("Failed to get assignment of " + topologyId);
continue;
}
assignments.put(topologyId, assignment);
}
Map<Integer, WorkerSummary> portWorkerSummarys = new TreeMap<Integer, WorkerSummary>();
for (Entry<String, Assignment> entry : assignments.entrySet()) {
String topologyId = entry.getKey();
Assignment assignment = entry.getValue();
Map<Integer, String> taskToComponent =
Cluster.topology_task_info(stormClusterState, topologyId);
Map<Integer, ResourceAssignment> taskToResource = assignment.getTaskToResource();
for (Entry<Integer, ResourceAssignment> resourceEntry : taskToResource.entrySet()) {
Integer taskId = resourceEntry.getKey();
ResourceAssignment resourceAssignment = resourceEntry.getValue();
if (supervisorId.equals(resourceAssignment.getSupervisorId()) == false) {
continue;
}
supervisorInfo.allocResource(resourceAssignment);
Integer port = resourceAssignment.getPort();
WorkerSummary workerSummary = portWorkerSummarys.get(port);
if (workerSummary == null) {
workerSummary = new WorkerSummary();
workerSummary.set_port(port);
workerSummary.set_topology(topologyId);
workerSummary.set_tasks(new ArrayList<TaskSummary>());
portWorkerSummarys.put(port, workerSummary);
}
String componentName = taskToComponent.get(taskId);
int uptime = TimeUtils.time_delta(assignment.getTaskStartTimeSecs().get(taskId));
List<TaskSummary> tasks = workerSummary.get_tasks();
TaskSummary taskSummary =
NimbusUtils.mkSimpleTaskSummary(
resourceAssignment, taskId, componentName, host, uptime);
tasks.add(taskSummary);
}
}
List<WorkerSummary> wokersList = new ArrayList<WorkerSummary>();
wokersList.addAll(portWorkerSummarys.values());
SupervisorSummary supervisorSummary =
NimbusUtils.mkSupervisorSummary(supervisorInfo, supervisorId);
return new SupervisorWorkers(supervisorSummary, wokersList);
} catch (TException e) {
LOG.info("Failed to get ClusterSummary ", e);
throw e;
} catch (Exception e) {
LOG.info("Failed to get ClusterSummary ", e);
throw new TException(e);
}
}
/**
* get cluster's summary, it will contain SupervisorSummary and TopologySummary
*
* @return ClusterSummary
*/
@Override
public ClusterSummary getClusterInfo() throws TException {
try {
StormClusterState stormClusterState = data.getStormClusterState();
Map<String, Assignment> assignments = new HashMap<String, Assignment>();
// get nimbus running time
int uptime = data.uptime();
// get TopologySummary
List<TopologySummary> topologySummaries = new ArrayList<TopologySummary>();
// get all active topology's StormBase
Map<String, StormBase> bases = Cluster.topology_bases(stormClusterState);
for (Entry<String, StormBase> entry : bases.entrySet()) {
String topologyId = entry.getKey();
StormBase base = entry.getValue();
Assignment assignment = stormClusterState.assignment_info(topologyId, null);
if (assignment == null) {
LOG.error("Failed to get assignment of " + topologyId);
continue;
}
assignments.put(topologyId, assignment);
String group = "default";
if (data.isGroupMode()) group = base.getGroup();
if (group == null) group = "default";
TopologySummary topology =
NimbusUtils.mkTopologySummary(
assignment,
topologyId,
base.getStormName(),
base.getStatusString(),
TimeUtils.time_delta(base.getLanchTimeSecs()),
group);
topologySummaries.add(topology);
}
// all supervisors
Map<String, SupervisorInfo> supervisorInfos =
Cluster.allSupervisorInfo(stormClusterState, null);
// generate SupervisorSummaries
List<SupervisorSummary> supervisorSummaries =
NimbusUtils.mkSupervisorSummaries(supervisorInfos, assignments);
return new ClusterSummary(
supervisorSummaries,
uptime,
topologySummaries,
data.getGroupToTopology(),
data.getGroupToResource(),
data.getGroupToUsedResource(),
data.isGroupMode());
} catch (TException e) {
LOG.info("Failed to get ClusterSummary ", e);
throw e;
} catch (Exception e) {
LOG.info("Failed to get ClusterSummary ", e);
throw new TException(e);
}
}
/**
* Submit one Topology
*
* @param topologyname String: topology name
* @param uploadedJarLocation String: already uploaded jar path
* @param jsonConf String: jsonConf serialize all toplogy configuration to Json
* @param topology StormTopology: topology Object
*/
@SuppressWarnings("unchecked")
@Override
public void submitTopologyWithOpts(
String topologyname,
String uploadedJarLocation,
String jsonConf,
StormTopology topology,
SubmitOptions options)
throws AlreadyAliveException, InvalidTopologyException, TopologyAssignException, TException {
LOG.info("Receive " + topologyname + ", uploadedJarLocation:" + uploadedJarLocation);
// @@@ Move validate topologyname in client code
try {
checkTopologyActive(data, topologyname, false);
} catch (AlreadyAliveException e) {
LOG.info(topologyname + " is already exist ");
throw e;
} catch (Exception e) {
LOG.info("Failed to check whether topology is alive or not", e);
throw new TException(e);
}
int counter = data.getSubmittedCount().incrementAndGet();
String topologyId = topologyname + "-" + counter + "-" + TimeUtils.current_time_secs();
Map<Object, Object> serializedConf = (Map<Object, Object>) JStormUtils.from_json(jsonConf);
if (serializedConf == null) {
LOG.warn("Failed to serialized Configuration");
throw new InvalidTopologyException("Failed to serilaze topology configuration");
}
serializedConf.put(Config.TOPOLOGY_ID, topologyId);
serializedConf.put(Config.TOPOLOGY_NAME, topologyname);
try {
Map<Object, Object> stormConf;
stormConf = NimbusUtils.normalizeConf(conf, serializedConf, topology);
Map<Object, Object> totalStormConf = new HashMap<Object, Object>(conf);
totalStormConf.putAll(stormConf);
StormTopology normalizedTopology = NimbusUtils.normalizeTopology(stormConf, topology);
// this validates the structure of the topology
Common.validate_basic(normalizedTopology, totalStormConf, topologyId);
// don't need generate real topology, so skip Common.system_topology
// Common.system_topology(totalStormConf, topology);
StormClusterState stormClusterState = data.getStormClusterState();
// create /local-dir/nimbus/topologyId/xxxx files
setupStormCode(conf, topologyId, uploadedJarLocation, stormConf, normalizedTopology);
// generate TaskInfo for every bolt or spout in ZK
// /ZK/tasks/topoologyId/xxx
setupZkTaskInfo(conf, topologyId, stormClusterState);
// make assignments for a topology
TopologyAssignEvent assignEvent = new TopologyAssignEvent();
assignEvent.setTopologyId(topologyId);
assignEvent.setScratch(false);
assignEvent.setTopologyName(topologyname);
assignEvent.setOldStatus(
Thrift.topologyInitialStatusToStormStatus(options.get_initial_status()));
TopologyAssign.push(assignEvent);
LOG.info("Submit for " + topologyname + " with conf " + serializedConf);
boolean isSuccess = assignEvent.waitFinish();
if (isSuccess == true) {
LOG.info("Finish submit for " + topologyname);
} else {
throw new FailedAssignTopologyException(assignEvent.getErrorMsg());
}
} catch (FailedAssignTopologyException e) {
StringBuilder sb = new StringBuilder();
sb.append("Fail to sumbit topology, Root cause:");
if (e.getMessage() == null) {
sb.append("submit timeout");
} else {
sb.append(e.getMessage());
}
sb.append("\n\n");
sb.append("topologyId:" + topologyId);
sb.append(", uploadedJarLocation:" + uploadedJarLocation + "\n");
LOG.error(sb.toString(), e);
throw new TopologyAssignException(sb.toString());
} catch (InvalidParameterException e) {
StringBuilder sb = new StringBuilder();
sb.append("Fail to sumbit topology ");
sb.append(e.getMessage());
sb.append(", cause:" + e.getCause());
sb.append("\n\n");
sb.append("topologyId:" + topologyId);
sb.append(", uploadedJarLocation:" + uploadedJarLocation + "\n");
LOG.error(sb.toString(), e);
throw new InvalidParameterException(sb.toString());
} catch (Throwable e) {
StringBuilder sb = new StringBuilder();
sb.append("Fail to sumbit topology ");
sb.append(e.getMessage());
sb.append(", cause:" + e.getCause());
sb.append("\n\n");
sb.append("topologyId:" + topologyId);
sb.append(", uploadedJarLocation:" + uploadedJarLocation + "\n");
LOG.error(sb.toString(), e);
throw new TopologyAssignException(sb.toString());
}
}