@Override
public void run() {
checkShutDown();
if (WorkerData.instance.nextBeat()) {
try {
TransactionContext.push();
doLaundryList(Singletons.getCoordinator());
} catch (Exception e) {
e.printStackTrace();
} finally {
TransactionContext.pop();
}
}
System.gc();
}
public void doLaundryList(Coordiantor coordinator) throws Exception {
/* if configuration was updated and load all updates
*/
WorkerData.instance.checkGlobalConfig(coordinator);
/* Check if host object initialized and if not make sure that
* host/group exists in db */
coordinator.ensureHost(WorkerData.instance);
/* Check CPU & Memory and up update
* statistics */
agent.runStatistics();
coordinator.storeStatistics(WorkerData.instance);
/* Check all active workers in the same group,
* calculate next beat time
* Update status of the workers and beet time
* on the same machine if necessary.
*/
coordinator.ensureWorker(WorkerData.instance);
Map<String, DomainTemplate> activeDomains = Singletons.getGlobalConfig().activeDomains();
List<BeeQueueDomain> domains = LazyList.morph(BeeQueueDomain.TO_DOMAIN, activeDomains.keySet());
coordinator.ensureDomains(domains);
/* Process all messages and create stages, create runs for
* unblocked stages. Pick runs to execute. Execute.
*/
coordinator.processEmittedMessages();
BeeQueueStage readyToRun = coordinator.pickStageToRun();
if (readyToRun != null) {
RunHelper runHelper = new RunHelper(readyToRun, WorkerData.instance.worker);
coordinator.storeRun(runHelper.run);
runHelper.start();
coordinator.storeRun(runHelper.run);
coordinator.updateStage(runHelper.stage);
}
/* Run ps. Identify all processes that currently executed.
* Check all processes that finished on host an update ther stages
* appropriately
*/
Map<Long, BeeQueueRun> notFinishedRuns = new LinkedHashMap<Long, BeeQueueRun>();
for (BeeQueueRun run : coordinator.allCurrentRuns(WorkerData.instance)) {
File rc = new File(run.dir, "rc.log");
if (rc.exists()) {
int returnCode = Integer.parseInt(Files.readAll(rc).trim());
run.state = returnCode == 0 ? RunState.SUCCESS : RunState.FAILURE;
coordinator.storeRun(run);
if (run.state == RunState.SUCCESS) {
run.stage.newState = StageState.SUCCESS;
} else {
run.stage.newState = run.stage.retriesLeft > 0 ? StageState.READY : StageState.FAILURE;
}
coordinator.updateStage(run.stage);
} else {
notFinishedRuns.put(run.id, run);
}
}
ProcRawData[] ps = agent.runProcessStatistics();
List<BeeQueueProcess> allActiveProcesses =
coordinator.allActiveProcessesOnHost(WorkerData.instance.host);
for (int i = 0; i < ps.length; i++) {
ProcRawData procRawData = ps[i];
if (procRawData.env != null && procRawData.env.get(RunHelper.BQ_RUN_ID) != null) {
BeeQueueProcess process = new BeeQueueProcess();
process.runId = Long.parseLong((String) procRawData.env.get(RunHelper.BQ_RUN_ID));
process.pid = "" + procRawData.pid;
process.ppid = "" + procRawData.state.getPpid();
process.processName = procRawData.allArgs();
coordinator.storeProcess(process);
allActiveProcesses.remove(process);
BeeQueueRun runThatStillHaveProcesses = notFinishedRuns.remove(process.runId);
if (runThatStillHaveProcesses != null) {
runThatStillHaveProcesses.justUpTimeStamp = true;
coordinator.storeRun(runThatStillHaveProcesses);
}
}
}
for (BeeQueueProcess process : allActiveProcesses) {
process.down = true;
coordinator.storeProcess(process);
}
if (notFinishedRuns.size() > 0) {
ToStringUtil.out("Cannot detect any info about processes:\n", notFinishedRuns);
}
coordinator.updateFinishedJobsAndMessages();
}