Examples with AzureSparkCosmosCluster com.microsoft.azure.hdinsight.sdk.common.azure.serverless.AzureSparkCosmosCluster used on opensource projects

Example 1 with AzureSparkCosmosCluster

use of com.microsoft.azure.hdinsight.sdk.common.azure.serverless.AzureSparkCosmosCluster in project azure-tools-for-java by Microsoft.

the class SparkBatchJobDeployFactory method buildSparkBatchJobDeploy.

public Deployable buildSparkBatchJobDeploy(@NotNull final SparkSubmitModel submitModel, @NotNull final IClusterDetail clusterDetail) throws ExecutionException {
    // get storage account and access token from submitModel
    final IHDIStorageAccount storageAccount;
    final String accessToken;
    final String accessKey;
    final String destinationRootPath;
    HttpObservable httpObservable = null;
    Deployable jobDeploy = null;
    final String clusterName = submitModel.getSubmissionParameter().getClusterName();
    final SparkSubmitStorageType storageAcccountType = submitModel.getJobUploadStorageModel().getStorageAccountType();
    if (storageAcccountType == null) {
        throw new ExecutionException("Job upload storage account type is empty.");
    }
    final String subscription = submitModel.getJobUploadStorageModel().getSelectedSubscription();
    // For HDI Reader cluster, Ambari credential is necessary for job submission.
    if (ClusterManagerEx.getInstance().isHdiReaderCluster(clusterDetail)) {
        try {
            if (clusterDetail.getHttpUserName() == null || clusterDetail.getHttpPassword() == null) {
                throw new ExecutionException("No Ambari permission to submit job to the selected cluster");
            }
        } catch (final HDIException ex) {
            log().warn("Error getting cluster credential. Cluster Name: " + clusterName);
            log().warn(ExceptionUtils.getStackTrace(ex));
            throw new ExecutionException("Error getting Ambari credential for this cluster");
        }
    }
    switch(storageAcccountType) {
        case BLOB:
            final String storageAccountName = submitModel.getJobUploadStorageModel().getStorageAccount();
            if (StringUtils.isBlank(storageAccountName)) {
                throw new ExecutionException("Can't get the valid storage account");
            }
            final String fullStorageBlobName = ClusterManagerEx.getInstance().getBlobFullName(storageAccountName);
            final String key = submitModel.getJobUploadStorageModel().getStorageKey();
            final String container = submitModel.getJobUploadStorageModel().getSelectedContainer();
            if (StringUtils.isBlank(key) || StringUtils.isBlank(container)) {
                throw new ExecutionException("Can't get the valid key or container name");
            }
            storageAccount = new HDStorageAccount(clusterDetail, fullStorageBlobName, key, false, container);
            jobDeploy = new LegacySDKDeploy(storageAccount);
            break;
        case DEFAULT_STORAGE_ACCOUNT:
            try {
                clusterDetail.getConfigurationInfo();
                storageAccount = clusterDetail.getStorageAccount();
                if (storageAccount == null) {
                    final String errorMsg = "Cannot get storage account from cluster";
                    log().warn(String.format("%s. Cluster: %s.", errorMsg, clusterDetail.getName()));
                    throw new ExecutionException(errorMsg);
                } else if (storageAccount.getAccountType() == StorageAccountType.ADLSGen2) {
                    final String defaultStorageRootPath = clusterDetail.getDefaultStorageRootPath();
                    if (defaultStorageRootPath == null) {
                        throw new ExecutionException(String.format("cluster %s's default storage root path is empty", clusterDetail.getName()));
                    }
                    final URI rawDestinationRootURI = UriUtil.normalizeWithSlashEnding(URI.create(defaultStorageRootPath)).resolve(SparkSubmissionContentPanel.Constants.submissionFolder + "/");
                    destinationRootPath = AbfsUri.parse(rawDestinationRootURI.toString()).getUrl().toString();
                    if (clusterDetail instanceof AzureAdAccountDetail) {
                        httpObservable = new ADLSGen2OAuthHttpObservable(((AzureAdAccountDetail) clusterDetail).getTenantId());
                    } else {
                        accessKey = ((ADLSGen2StorageAccount) storageAccount).getPrimaryKey();
                        httpObservable = new SharedKeyHttpObservable(storageAccount.getName(), accessKey);
                        if (StringUtils.isBlank(accessKey)) {
                            throw new ExecutionException("Cannot get valid access key for storage account");
                        }
                    }
                    jobDeploy = new ADLSGen2Deploy(httpObservable, destinationRootPath);
                } else if (storageAccount.getAccountType() == StorageAccountType.BLOB || storageAccount.getAccountType() == StorageAccountType.ADLS) {
                    if (clusterDetail instanceof SynapseCosmosSparkPool || clusterDetail instanceof AzureSparkCosmosCluster) {
                        final AzureHttpObservable http = clusterDetail instanceof SynapseCosmosSparkPool ? ((SynapseCosmosSparkPool) clusterDetail).getHttp() : ((AzureSparkCosmosCluster) clusterDetail).getHttp();
                        if (http == null) {
                            final String errorMsg = "Error preparing access token for ADLS Gen1 storage account";
                            log().warn(String.format("%s. Cluster: %s. Storage account: %s.", errorMsg, clusterDetail.getName(), storageAccount.getName()));
                            throw new ExecutionException("Error preparing access token for ADLS Gen1 storage account");
                        }
                        final String defaultStorageRootPath = clusterDetail.getDefaultStorageRootPath();
                        if (StringUtils.isBlank(defaultStorageRootPath)) {
                            final String errorMsg = "Error getting default storage root path for ADLS Gen1 storage account";
                            log().warn(String.format("%s. Cluster: %s. Storage account: %s.", errorMsg, clusterDetail.getName(), storageAccount.getName()));
                            throw new ExecutionException(errorMsg);
                        }
                        jobDeploy = new AdlsDeploy(defaultStorageRootPath, http.getAccessToken());
                    } else {
                        jobDeploy = new LegacySDKDeploy(storageAccount);
                    }
                }
            } catch (final Exception ex) {
                log().warn("Error getting cluster storage configuration. Error: " + ExceptionUtils.getStackTrace(ex));
            }
            break;
        case SPARK_INTERACTIVE_SESSION:
            jobDeploy = new LivySessionDeploy(clusterName);
            break;
        case ADLS_GEN1:
            final String rawRootPath = submitModel.getJobUploadStorageModel().getAdlsRootPath();
            if (StringUtils.isBlank(rawRootPath) || !AdlUri.isType(rawRootPath)) {
                throw new ExecutionException("Invalid adls root path input");
            }
            destinationRootPath = rawRootPath.endsWith("/") ? rawRootPath : rawRootPath + "/";
            // e.g. for adl://john.azuredatalakestore.net/root/path, adlsAccountName is john
            final String adlsAccountName = destinationRootPath.split("\\.")[0].split("//")[1];
            Optional<SubscriptionDetail> subscriptionDetail = Optional.empty();
            try {
                subscriptionDetail = AuthMethodManager.getInstance().getAzureManager().getSubscriptionManager().getSelectedSubscriptionDetails().stream().filter((detail) -> detail.getSubscriptionName().equals(subscription)).findFirst();
            } catch (final Exception ignore) {
            }
            if (!subscriptionDetail.isPresent()) {
                throw new ExecutionException("Error getting subscription info. Please select correct subscription");
            }
            // get Access Token
            try {
                accessToken = AzureSparkClusterManager.getInstance().getAccessToken(subscriptionDetail.get().getTenantId());
            } catch (final IOException ex) {
                log().warn("Error getting access token based on the given ADLS root path. " + ExceptionUtils.getStackTrace(ex));
                throw new ExecutionException("Error getting access token based on the given ADLS root path");
            }
            jobDeploy = new AdlsDeploy(destinationRootPath, accessToken);
            break;
        case ADLS_GEN2_FOR_OAUTH:
        case ADLS_GEN2:
            destinationRootPath = submitModel.getJobUploadStorageModel().getUploadPath();
            if (!AbfsUri.isType(destinationRootPath)) {
                throw new ExecutionException("Invalid ADLS GEN2 root path: " + destinationRootPath);
            }
            final AbfsUri destinationUri = AbfsUri.parse(destinationRootPath);
            if (clusterDetail instanceof AzureAdAccountDetail) {
                httpObservable = new ADLSGen2OAuthHttpObservable(((AzureAdAccountDetail) clusterDetail).getTenantId());
            } else {
                accessKey = submitModel.getJobUploadStorageModel().getAccessKey();
                if (StringUtils.isBlank(accessKey)) {
                    throw new ExecutionException("Invalid access key input");
                }
                final String accountName = destinationUri.getAccountName();
                httpObservable = new SharedKeyHttpObservable(accountName, accessKey);
            }
            jobDeploy = new ADLSGen2Deploy(httpObservable, destinationUri.getUrl().toString());
            break;
        case WEBHDFS:
            destinationRootPath = submitModel.getJobUploadStorageModel().getUploadPath();
            if (StringUtils.isBlank(destinationRootPath) || !destinationRootPath.matches(SparkBatchJob.WebHDFSPathPattern)) {
                throw new ExecutionException("Invalid webhdfs root path input");
            }
            // create httpobservable and jobDeploy
            try {
                if (clusterDetail instanceof ClusterDetail) {
                    httpObservable = new AzureHttpObservable(clusterDetail.getSubscription().getTenantId(), ApiVersion.VERSION);
                    jobDeploy = clusterDetail.getStorageAccount() != null && clusterDetail.getStorageAccount().getAccountType() == StorageAccountType.ADLS ? new ADLSGen1HDFSDeploy(clusterDetail, httpObservable, destinationRootPath) : null;
                } else if (clusterDetail instanceof SqlBigDataLivyLinkClusterDetail) {
                    httpObservable = new HttpObservable(clusterDetail.getHttpUserName(), clusterDetail.getHttpPassword());
                    jobDeploy = new WebHDFSDeploy(clusterDetail, httpObservable, destinationRootPath);
                }
            } catch (final HDIException ignore) {
            }
            if (httpObservable == null || jobDeploy == null) {
                throw new ExecutionException("Error preparing webhdfs uploading info based on the given cluster");
            }
            break;
        default:
            throw new ExecutionException("Unsupported job upload storage type");
    }
    // TODO:use httpobservable to replace sparkbathsubmission and deprecate the old constructor.
    return jobDeploy;
}