/**
* make assignments for a topology The nimbus core function, this function has been totally
* rewrite
*
* @param nimbusData NimbusData
* @param topologyId String
* @param isScratch Boolean: isScratch is false unless rebalancing the topology
* @throws Exception
*/
public Assignment mkAssignment(TopologyAssignEvent event) throws Exception {
String topologyId = event.getTopologyId();
LOG.info("Determining assignment for " + topologyId);
TopologyAssignContext context = prepareTopologyAssign(event);
Set<ResourceWorkerSlot> assignments = null;
if (!StormConfig.local_mode(nimbusData.getConf())) {
IToplogyScheduler scheduler = schedulers.get(DEFAULT_SCHEDULER_NAME);
assignments = scheduler.assignTasks(context);
} else {
assignments = mkLocalAssignment(context);
}
Assignment assignment = null;
Map<String, String> nodeHost =
getTopologyNodeHost(context.getCluster(), context.getOldAssignment(), assignments);
Map<Integer, Integer> startTimes =
getTaskStartTimes(context, nimbusData, topologyId, context.getOldAssignment(), assignments);
String codeDir = StormConfig.masterStormdistRoot(nimbusData.getConf(), topologyId);
assignment = new Assignment(codeDir, assignments, nodeHost, startTimes);
StormClusterState stormClusterState = nimbusData.getStormClusterState();
stormClusterState.set_assignment(topologyId, assignment);
// update task heartbeat's start time
NimbusUtils.updateTaskHbStartTime(nimbusData, assignment, topologyId);
// Update metrics information in ZK when rebalance or reassignment
// Only update metrics monitor status when creating topology
if (context.getAssignType() == TopologyAssignContext.ASSIGN_TYPE_REBALANCE
|| context.getAssignType() == TopologyAssignContext.ASSIGN_TYPE_MONITOR)
NimbusUtils.updateMetricsInfo(nimbusData, topologyId, assignment);
else metricsMonitor(event);
LOG.info("Successfully make assignment for topology id " + topologyId + ": " + assignment);
return assignment;
}
/**
* set topology status as active
*
* @param topologyname
*/
@Override
public void activate(String topologyName) throws NotAliveException, TException {
try {
NimbusUtils.transitionName(data, topologyName, true, StatusType.activate);
} catch (NotAliveException e) {
String errMsg = "Activate Error, no this topology " + topologyName;
LOG.error(errMsg, e);
throw new NotAliveException(errMsg);
} catch (Exception e) {
String errMsg = "Failed to active topology " + topologyName;
LOG.error(errMsg, e);
throw new TException(errMsg);
}
}
/**
* find all alived taskid Does not assume that clocks are synchronized. Task heartbeat is only
* used so that nimbus knows when it's received a new heartbeat. All timing is done by nimbus and
* tracked through task-heartbeat-cache 获取topology下的所有alive的task
*
* @param conf
* @param topologyId
* @param stormClusterState
* @param taskIds
* @param taskStartTimes
* @param taskHeartbeatsCache --Map<topologyId, Map<taskid, Map<tkHbCacheTime, time>>>
* @return Set<Integer> : taskid
* @throws Exception
*/
public Set<Integer> getAliveTasks(String topologyId, Set<Integer> taskIds) throws Exception {
Set<Integer> aliveTasks = new HashSet<Integer>();
// taskIds is the list from ZK /ZK-DIR/tasks/topologyId
for (int taskId : taskIds) {
boolean isDead = NimbusUtils.isTaskDead(nimbusData, topologyId, taskId);
if (isDead == false) {
aliveTasks.add(taskId);
}
}
return aliveTasks;
}
/**
* 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 void killTopologyWithOpts(String topologyName, KillOptions options)
throws NotAliveException, TException {
try {
checkTopologyActive(data, topologyName, true);
Integer wait_amt = null;
if (options.is_set_wait_secs()) {
wait_amt = options.get_wait_secs();
}
NimbusUtils.transitionName(data, topologyName, true, StatusType.kill, wait_amt);
} catch (NotAliveException e) {
String errMsg = "KillTopology Error, no this topology " + topologyName;
LOG.error(errMsg, e);
throw new NotAliveException(errMsg);
} catch (Exception e) {
String errMsg = "Failed to kill topology " + topologyName;
LOG.error(errMsg, e);
throw new TException(errMsg);
}
}
/**
* rebalance one topology @@@ rebalance options hasn't implements
*
* <p>It is used to let workers wait several seconds to finish jobs
*
* @param topologyname String
* @param options RebalanceOptions
*/
@Override
public void rebalance(String topologyName, RebalanceOptions options)
throws NotAliveException, TException, InvalidTopologyException {
try {
checkTopologyActive(data, topologyName, true);
Integer wait_amt = null;
if (options != null && options.is_set_wait_secs()) {
wait_amt = options.get_wait_secs();
}
NimbusUtils.transitionName(data, topologyName, true, StatusType.rebalance, wait_amt);
} catch (NotAliveException e) {
String errMsg = "Rebalance Error, no this topology " + topologyName;
LOG.error(errMsg, e);
throw new NotAliveException(errMsg);
} catch (Exception e) {
String errMsg = "Failed to rebalance topology " + topologyName;
LOG.error(errMsg, e);
throw new TException(errMsg);
}
}
@Override
public <T> Object execute(T... args) {
boolean isSetTaskInfo = false;
try {
Boolean reassign = (Boolean) args[1];
Map<Object, Object> conf = (Map<Object, Object>) args[2]; // args[0]:
// delay,
// args[1]:
// reassign_flag,
// args[2]:
// conf
if (conf != null) {
boolean isConfUpdate = false;
Map stormConf = data.getConf();
// Update topology code
Map topoConf = StormConfig.read_nimbus_topology_conf(stormConf, topologyid);
StormTopology rawOldTopology = StormConfig.read_nimbus_topology_code(stormConf, topologyid);
StormTopology rawNewTopology = NimbusUtils.normalizeTopology(conf, rawOldTopology, true);
StormTopology sysOldTopology = rawOldTopology.deepCopy();
StormTopology sysNewTopology = rawNewTopology.deepCopy();
if (conf.get(Config.TOPOLOGY_ACKER_EXECUTORS) != null) {
Common.add_acker(topoConf, sysOldTopology);
Common.add_acker(conf, sysNewTopology);
int ackerNum = JStormUtils.parseInt(conf.get(Config.TOPOLOGY_ACKER_EXECUTORS));
int oldAckerNum = JStormUtils.parseInt(topoConf.get(Config.TOPOLOGY_ACKER_EXECUTORS));
LOG.info("Update acker from oldAckerNum=" + oldAckerNum + " to ackerNum=" + ackerNum);
topoConf.put(Config.TOPOLOGY_ACKER_EXECUTORS, ackerNum);
isConfUpdate = true;
}
// If scale-out, setup task info for new added tasks
setTaskInfo(sysOldTopology, sysNewTopology);
isSetTaskInfo = true;
// If everything is OK, write topology code into disk
StormConfig.write_nimbus_topology_code(
stormConf, topologyid, Utils.serialize(rawNewTopology));
// Update topology conf if worker num has been updated
Set<Object> keys = conf.keySet();
Integer workerNum = JStormUtils.parseInt(conf.get(Config.TOPOLOGY_WORKERS));
if (workerNum != null) {
Integer oldWorkerNum = JStormUtils.parseInt(topoConf.get(Config.TOPOLOGY_WORKERS));
topoConf.put(Config.TOPOLOGY_WORKERS, workerNum);
isConfUpdate = true;
LOG.info("Update worker num from " + oldWorkerNum + " to " + workerNum);
}
if (keys.contains(Config.ISOLATION_SCHEDULER_MACHINES)) {
topoConf.put(
Config.ISOLATION_SCHEDULER_MACHINES, conf.get(Config.ISOLATION_SCHEDULER_MACHINES));
}
if (isConfUpdate) {
StormConfig.write_nimbus_topology_conf(stormConf, topologyid, topoConf);
}
}
TopologyAssignEvent event = new TopologyAssignEvent();
event.setTopologyId(topologyid);
event.setScratch(true);
event.setOldStatus(oldStatus);
event.setReassign(reassign);
if (conf != null) event.setScaleTopology(true);
TopologyAssign.push(event);
event.waitFinish();
} catch (Exception e) {
LOG.error("do-rebalance error!", e);
// Rollback the changes on ZK
if (isSetTaskInfo) {
try {
StormClusterState clusterState = data.getStormClusterState();
clusterState.remove_task(topologyid, newTasks);
} catch (Exception e1) {
LOG.error("Failed to rollback the changes on ZK for task-" + newTasks, e);
}
}
}
DelayStatusTransitionCallback delayCallback =
new DelayStatusTransitionCallback(
data, topologyid, oldStatus, StatusType.rebalancing, StatusType.done_rebalance);
return delayCallback.execute();
}
@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());
}
}