Examples with CLUSTER_SCALING_STOPSTART_UPSCALE_CMHOSTSSTARTED com.sequenceiq.cloudbreak.event.ResourceEvent.CLUSTER_SCALING_STOPSTART_UPSCALE_CMHOSTSSTARTED used on opensource projects

Example 1 with CLUSTER_SCALING_STOPSTART_UPSCALE_CMHOSTSSTARTED

use of com.sequenceiq.cloudbreak.event.ResourceEvent.CLUSTER_SCALING_STOPSTART_UPSCALE_CMHOSTSSTARTED in project cloudbreak by hortonworks.

the class StopStartUpscaleCommissionViaCMHandler method doAccept.

@Override
protected Selectable doAccept(HandlerEvent<StopStartUpscaleCommissionViaCMRequest> event) {
    StopStartUpscaleCommissionViaCMRequest request = event.getData();
    LOGGER.info("StopStartUpscaleCommissionViaCMHandler for: {}, {}", event.getData().getResourceId(), event);
    LOGGER.debug("StartedInstancesToCommission: {}, servicesNotRunningInstancesToCommission: {}", request.getStartedInstancesToCommission(), request.getServicesNotRunningInstancesToCommission());
    List<InstanceMetaData> allInstancesToCommission = new LinkedList<>();
    allInstancesToCommission.addAll(request.getStartedInstancesToCommission());
    allInstancesToCommission.addAll(request.getServicesNotRunningInstancesToCommission());
    try {
        Stack stack = stackService.getByIdWithLists(request.getResourceId());
        Cluster cluster = stack.getCluster();
        flowMessageService.fireEventAndLog(stack.getId(), UPDATE_IN_PROGRESS.name(), CLUSTER_SCALING_STOPSTART_UPSCALE_WAITING_HOSTSTART, String.valueOf(allInstancesToCommission.size()));
        ClusterSetupService clusterSetupService = clusterApiConnectors.getConnector(stack).clusterSetupService();
        clusterSetupService.waitForHostsHealthy(new HashSet<>(allInstancesToCommission));
        flowMessageService.fireEventAndLog(stack.getId(), UPDATE_IN_PROGRESS.name(), CLUSTER_SCALING_STOPSTART_UPSCALE_CMHOSTSSTARTED, String.valueOf(allInstancesToCommission.size()));
        ClusterCommissionService clusterCommissionService = clusterApiConnectors.getConnector(stack).clusterCommissionService();
        Set<String> hostNames = allInstancesToCommission.stream().map(i -> i.getDiscoveryFQDN()).collect(Collectors.toSet());
        LOGGER.debug("HostNames to recommission: count={}, hostNames={}", hostNames.size(), hostNames);
        HostGroup hostGroup = hostGroupService.getByClusterIdAndName(cluster.getId(), request.getHostGroupName()).orElseThrow(NotFoundException.notFound("hostgroup", request.getHostGroupName()));
        Map<String, InstanceMetaData> hostsToRecommission = clusterCommissionService.collectHostsToCommission(hostGroup, hostNames);
        List<String> missingHostsInCm = Collections.emptyList();
        if (hostNames.size() != hostsToRecommission.size()) {
            missingHostsInCm = hostNames.stream().filter(h -> !hostsToRecommission.containsKey(h)).collect(Collectors.toList());
            LOGGER.info("Found fewer instances in CM to commission, as compared to initial ask. foundCount={}, initialCount={}, missingHostsInCm={}", hostsToRecommission.size(), hostNames.size(), missingHostsInCm);
        }
        // TODO CB-15132: Eventually ensure CM, relevant services (YARN RM) are in a functional state - or fail/delay the operation
        // TODO CB-15132: Potentially poll nodes for success. Don't fail the entire operation if a single node fails to commission.
        // What would need to happen to the CM command in this case? (Can only work in the presence of a co-operative CM API call.
        // Alternately this could go straight to the service)
        Set<String> recommissionedHostnames = Collections.emptySet();
        if (hostsToRecommission.size() > 0) {
            recommissionedHostnames = clusterCommissionService.recommissionClusterNodes(hostsToRecommission);
        // TODO CB-15132: Maybe wait for services to start / force CM sync.
        }
        List<String> allMissingRecommissionHostnames = null;
        if (missingHostsInCm.size() > 0) {
            allMissingRecommissionHostnames = new LinkedList<>(missingHostsInCm);
        }
        if (hostsToRecommission.size() != recommissionedHostnames.size()) {
            Set<String> finalRecommissionedHostnames = recommissionedHostnames;
            List<String> additionalMissingRecommissionHostnames = hostsToRecommission.keySet().stream().filter(h -> !finalRecommissionedHostnames.contains(h)).collect(Collectors.toList());
            LOGGER.info("Recommissioned fewer instances than requested. recommissionedCount={}, expectedCount={}, initialCount={}, notRecommissioned=[{}]", recommissionedHostnames.size(), hostsToRecommission.size(), hostNames.size(), additionalMissingRecommissionHostnames);
            if (allMissingRecommissionHostnames == null) {
                allMissingRecommissionHostnames = new LinkedList<>();
            }
            allMissingRecommissionHostnames.addAll(additionalMissingRecommissionHostnames);
        }
        return new StopStartUpscaleCommissionViaCMResult(request, recommissionedHostnames, allMissingRecommissionHostnames);
    } catch (Exception e) {
        // TODO CB-15132: This can be improved based on where and when the Exception occurred to potentially rollback certain aspects.
        // ClusterClientInitException is one which is explicitly thrown.
        String message = "Failed while attempting to commission nodes via CM";
        LOGGER.error(message);
        return new StopStartUpscaleCommissionViaCMResult(message, e, request);
    }
}