[GEODE-8507] Calling gfsh connect while JMX manager is starting up can result in error

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.14.0
• Component/s: management
• Assignee: Unassigned
• Reporter: Dan Smith
• Created: 17/Sep/20 21:16
• Updated: 17/Sep/20 21:16

Description

Running `gfsh -e connect` on a locator at the wrong time results in this error:

Exception caused JMX Manager startup to fail because: 'Manager is already running'

Looking at where this message is generated, I think this is coming from this code in JmxLocatorManager that tries to find a JMX manager and start one if it is not running

//*** Manager may not be started at this point
 List<JmxManagerProfile> alreadyManaging =
        this.cache.getJmxManagerAdvisor().adviseAlreadyManaging();
    if (alreadyManaging.isEmpty()) {
      List<JmxManagerProfile> willingToManage =
          this.cache.getJmxManagerAdvisor().adviseWillingToManage();
      if (!willingToManage.isEmpty()) {
        synchronized (this) {
          alreadyManaging = this.cache.getJmxManagerAdvisor().adviseAlreadyManaging();
          if (alreadyManaging.isEmpty()) {
            willingToManage = this.cache.getJmxManagerAdvisor().adviseWillingToManage();
            if (!willingToManage.isEmpty()) {
              JmxManagerProfile p = willingToManage.get(0);
              if (p.getDistributedMember().equals(this.cache.getMyId())) {
                if (logger.isDebugEnabled()) {
                  logger.debug("Locator starting jmx manager in its JVM");
                }
                try {
//*** This call can throw an AlreadyRunningException if the manager is already running
//*** failing this whole method.
                  ManagementService.getManagementService(this.cache).startManager();
                } catch (CancelException ex) {
                  // ignore
                } catch (VirtualMachineError err) {
                  SystemFailure.initiateFailure(err);
                  // If this ever returns, rethrow the error. We're poisoned
                  // now, so don't let this thread continue.
                  throw err;
                } catch (Throwable t) {
                  SystemFailure.checkFailure();
                  return new JmxManagerLocatorResponse(null, 0, false, t);
                }