use of com.netflix.spinnaker.halyard.core.error.v1.HalException in project halyard by spinnaker.
the class RoscoProfileFactory method getImageProviders.
protected Providers getImageProviders(String version) {
InputStream is;
try {
is = profileRegistry.readProfile(getArtifact().getName(), version, "images.yml");
} catch (IOException e) {
throw new HalException(Problem.Severity.FATAL, "Unable to read images.yml for rosco: " + e.getMessage(), e);
}
Object obj = yamlParser.load(is);
return objectMapper.convertValue(obj, Providers.class);
}
use of com.netflix.spinnaker.halyard.core.error.v1.HalException in project halyard by spinnaker.
the class Profile method writeStagedFile.
public void writeStagedFile(String stagingPath) {
AtomicFileWriter writer = null;
Path path = Paths.get(stagingPath, name);
log.info("Writing profile to " + path.toString() + " with " + requiredFiles.size() + " required files");
try {
writer = new AtomicFileWriter(path);
writer.write(contents);
writer.commit();
} catch (IOException ioe) {
ioe.printStackTrace();
throw new HalException(Problem.Severity.FATAL, "Failed to write config for profile " + path.toFile().getName() + ": " + ioe.getMessage());
} finally {
if (writer != null) {
writer.close();
}
}
}
use of com.netflix.spinnaker.halyard.core.error.v1.HalException in project halyard by spinnaker.
the class DistributedDeployer method deploy.
@Override
public RemoteAction deploy(DistributedServiceProvider<T> serviceProvider, AccountDeploymentDetails<T> deploymentDetails, ResolvedConfiguration resolvedConfiguration, List<SpinnakerService.Type> serviceTypes) {
SpinnakerRuntimeSettings runtimeSettings = resolvedConfiguration.getRuntimeSettings();
DaemonTaskHandler.newStage("Deploying Spinnaker");
// First deploy all services not owned by Spinnaker
for (DistributedService distributedService : serviceProvider.getPrioritizedDistributedServices(serviceTypes)) {
SpinnakerService service = distributedService.getService();
ServiceSettings settings = resolvedConfiguration.getServiceSettings(service);
if (!settings.getEnabled() || settings.getSkipLifeCycleManagement()) {
continue;
}
DaemonTaskHandler.newStage("Determining status of " + distributedService.getServiceName());
boolean safeToUpdate = settings.getSafeToUpdate();
RunningServiceDetails runningServiceDetails = distributedService.getRunningServiceDetails(deploymentDetails, runtimeSettings);
if (distributedService.isRequiredToBootstrap() || !safeToUpdate) {
deployServiceManually(deploymentDetails, resolvedConfiguration, distributedService, safeToUpdate);
} else {
DaemonResponse.StaticRequestBuilder<Void> builder = new DaemonResponse.StaticRequestBuilder<>(() -> {
if (runningServiceDetails.getLatestEnabledVersion() == null) {
DaemonTaskHandler.newStage("Deploying " + distributedService.getServiceName() + " via provider API");
deployServiceManually(deploymentDetails, resolvedConfiguration, distributedService, safeToUpdate);
} else {
DaemonTaskHandler.newStage("Deploying " + distributedService.getServiceName() + " via red/black");
Orca orca = serviceProvider.getDeployableService(SpinnakerService.Type.ORCA_BOOTSTRAP, Orca.class).connectToPrimaryService(deploymentDetails, runtimeSettings);
deployServiceWithOrca(deploymentDetails, resolvedConfiguration, orca, distributedService);
}
return null;
});
DaemonTaskHandler.submitTask(builder::build, "Deploy " + distributedService.getServiceName());
}
}
DaemonTaskHandler.message("Waiting on deployments to complete");
DaemonTaskHandler.reduceChildren(null, (t1, t2) -> null, (t1, t2) -> null).getProblemSet().throwifSeverityExceeds(Problem.Severity.WARNING);
DistributedService<Orca, T> orca = serviceProvider.getDeployableService(SpinnakerService.Type.ORCA);
Set<Integer> unknownVersions = reapOrcaServerGroups(deploymentDetails, runtimeSettings, orca);
reapRoscoServerGroups(deploymentDetails, runtimeSettings, serviceProvider.getDeployableService(SpinnakerService.Type.ROSCO));
if (!unknownVersions.isEmpty()) {
String versions = String.join(", ", unknownVersions.stream().map(orca::getVersionedName).collect(Collectors.toList()));
throw new HalException(new ProblemBuilder(Problem.Severity.ERROR, "The following orca versions (" + versions + ") could not safely be drained of work.").setRemediation("Please make sure that no pipelines are running, and manually destroy the server groups at those versions.").build());
}
return new RemoteAction();
}
use of com.netflix.spinnaker.halyard.core.error.v1.HalException in project halyard by spinnaker.
the class DistributedDeployer method rollback.
@Override
public void rollback(DistributedServiceProvider<T> serviceProvider, AccountDeploymentDetails<T> deploymentDetails, SpinnakerRuntimeSettings runtimeSettings, List<SpinnakerService.Type> serviceTypes) {
DaemonTaskHandler.newStage("Checking if it is safe to roll back all services");
for (DistributedService distributedService : serviceProvider.getPrioritizedDistributedServices(serviceTypes)) {
SpinnakerService service = distributedService.getService();
ServiceSettings settings = runtimeSettings.getServiceSettings(service);
boolean safeToUpdate = settings.getSafeToUpdate();
if (!settings.getEnabled() || distributedService.isRequiredToBootstrap() || !safeToUpdate || settings.getSkipLifeCycleManagement()) {
continue;
}
RunningServiceDetails runningServiceDetails = distributedService.getRunningServiceDetails(deploymentDetails, runtimeSettings);
if (runningServiceDetails.getInstances().keySet().size() == 1) {
throw new HalException(Problem.Severity.FATAL, "Service " + service.getCanonicalName() + " has only one server group - there is nothing to rollback to.");
}
}
DaemonTaskHandler.newStage("Rolling back all updatable services");
for (DistributedService distributedService : serviceProvider.getPrioritizedDistributedServices(serviceTypes)) {
SpinnakerService service = distributedService.getService();
ServiceSettings settings = runtimeSettings.getServiceSettings(service);
if (!settings.getEnabled() || settings.getSkipLifeCycleManagement()) {
continue;
}
boolean safeToUpdate = settings.getSafeToUpdate();
if (distributedService.isRequiredToBootstrap() || !safeToUpdate) {
// Do nothing, the bootstrapping services should already be running, and the services that can't be updated
// having nothing to rollback to
} else {
DaemonResponse.StaticRequestBuilder<Void> builder = new DaemonResponse.StaticRequestBuilder<>(() -> {
Orca orca = serviceProvider.getDeployableService(SpinnakerService.Type.ORCA_BOOTSTRAP, Orca.class).connectToPrimaryService(deploymentDetails, runtimeSettings);
DaemonTaskHandler.message("Rolling back " + distributedService.getServiceName() + " via Spinnaker red/black");
rollbackService(deploymentDetails, orca, distributedService, runtimeSettings);
return null;
});
DaemonTaskHandler.submitTask(builder::build, "Rollback " + distributedService.getServiceName());
}
}
DaemonTaskHandler.message("Waiting on rollbacks to complete");
DaemonTaskHandler.reduceChildren(null, (t1, t2) -> null, (t1, t2) -> null).getProblemSet().throwifSeverityExceeds(Problem.Severity.WARNING);
}
use of com.netflix.spinnaker.halyard.core.error.v1.HalException in project halyard by spinnaker.
the class DistributedDeployer method reapRoscoServerGroups.
private <T extends Account> void reapRoscoServerGroups(AccountDeploymentDetails<T> details, SpinnakerRuntimeSettings runtimeSettings, DistributedService<Rosco, T> roscoService) {
if (runtimeSettings.getServiceSettings(roscoService.getService()).getSkipLifeCycleManagement()) {
return;
}
ServiceSettings roscoSettings = runtimeSettings.getServiceSettings(roscoService.getService());
Rosco.AllStatus allStatus;
try {
Rosco rosco = roscoService.connectToPrimaryService(details, runtimeSettings);
allStatus = rosco.getAllStatus();
} catch (RetrofitError e) {
boolean enabled = roscoSettings.getEnabled() != null && roscoSettings.getEnabled();
if (enabled) {
throw new HalException(Problem.Severity.FATAL, "Rosco is enabled, and no connection to rosco could be established: " + e, e);
}
Response response = e.getResponse();
if (response == null) {
throw new IllegalStateException("Unknown connection failure: " + e, e);
}
// 404 when the service couldn't be found, 503 when k8s couldn't establish a connection
if (response.getStatus() == 404 || response.getStatus() == 503) {
log.info("Rosco is not enabled, and there are no server groups to reap");
return;
} else {
throw new HalException(Problem.Severity.FATAL, "Rosco is not enabled, but couldn't be connected to for unknown reason: " + e, e);
}
}
RunningServiceDetails roscoDetails = roscoService.getRunningServiceDetails(details, runtimeSettings);
Set<String> activeInstances = new HashSet<>();
allStatus.getInstances().forEach((s, e) -> {
if (e.getStatus().equals(Rosco.Status.RUNNING)) {
String[] split = s.split("@");
if (split.length != 2) {
log.warn("Unsupported rosco status format");
return;
}
String instanceId = split[1];
activeInstances.add(instanceId);
}
});
Map<Integer, Integer> executionsByServerGroupVersion = new HashMap<>();
roscoDetails.getInstances().forEach((s, is) -> {
int count = is.stream().reduce(0, (c, i) -> c + (activeInstances.contains(i) ? 1 : 0), (a, b) -> a + b);
executionsByServerGroupVersion.put(s, count);
});
// Omit the last deployed roscos from being deleted, since they are kept around for rollbacks.
List<Integer> allRoscos = new ArrayList<>(executionsByServerGroupVersion.keySet());
allRoscos.sort(Integer::compareTo);
cleanupServerGroups(details, roscoService, roscoSettings, executionsByServerGroupVersion, allRoscos);
}
Aggregations