Search in sources :

Example 1 with YarnProjectsQuota

use of io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota in project hopsworks by logicalclocks.

the class JupyterService method startNotebookServer.

@POST
@Path("/start")
@Consumes(MediaType.APPLICATION_JSON)
@Produces(MediaType.APPLICATION_JSON)
@AllowedProjectRoles({ AllowedProjectRoles.DATA_OWNER, AllowedProjectRoles.DATA_SCIENTIST })
@JWTRequired(acceptedTokens = { Audience.API }, allowedUserRoles = { "HOPS_ADMIN", "HOPS_USER" })
public Response startNotebookServer(JupyterSettings jupyterSettings, @Context HttpServletRequest req, @Context SecurityContext sc, @Context UriInfo uriInfo) throws ProjectException, HopsSecurityException, ServiceException, GenericException, JobException {
    Users hopsworksUser = jWTHelper.getUserPrincipal(sc);
    String hdfsUser = hdfsUsersController.getHdfsUserName(project, hopsworksUser);
    // from in the front-end
    if (jupyterSettings.getUsers() == null) {
        jupyterSettings.setUsers(hopsworksUser);
    }
    if (project.getPaymentType().equals(PaymentType.PREPAID)) {
        YarnProjectsQuota projectQuota = yarnProjectsQuotaFacade.findByProjectName(project.getName());
        if (projectQuota == null || projectQuota.getQuotaRemaining() <= 0) {
            throw new ProjectException(RESTCodes.ProjectErrorCode.PROJECT_QUOTA_ERROR, Level.FINE);
        }
    }
    if (project.getPythonEnvironment() == null) {
        throw new ProjectException(RESTCodes.ProjectErrorCode.ANACONDA_NOT_ENABLED, Level.FINE);
    }
    if (jupyterSettings.getMode() == null) {
        // set default mode for jupyter if mode is null
        jupyterSettings.setMode(JupyterMode.JUPYTER_LAB);
    }
    // Jupyter Git works only for JupyterLab
    if (jupyterSettings.isGitBackend() && jupyterSettings.getMode().equals(JupyterMode.JUPYTER_CLASSIC)) {
        throw new ServiceException(RESTCodes.ServiceErrorCode.JUPYTER_START_ERROR, Level.FINE, "Git support available only in JupyterLab");
    }
    // Do not allow auto push on shutdown if api key is missing
    GitConfig gitConfig = jupyterSettings.getGitConfig();
    if (jupyterSettings.isGitBackend() && gitConfig.getShutdownAutoPush() && Strings.isNullOrEmpty(gitConfig.getApiKeyName())) {
        throw new ServiceException(RESTCodes.ServiceErrorCode.JUPYTER_START_ERROR, Level.FINE, "Auto push not supported if api key is not configured.");
    }
    // Verify that API token has got write access on the repo if ShutdownAutoPush is enabled
    if (jupyterSettings.isGitBackend() && gitConfig.getShutdownAutoPush() && !jupyterNbVCSController.hasWriteAccess(hopsworksUser, gitConfig.getApiKeyName(), gitConfig.getRemoteGitURL(), gitConfig.getGitBackend())) {
        throw new ServiceException(RESTCodes.ServiceErrorCode.JUPYTER_START_ERROR, Level.FINE, "API token " + gitConfig.getApiKeyName() + " does not have write access on " + gitConfig.getRemoteGitURL());
    }
    JupyterProject jp = jupyterFacade.findByUser(hdfsUser);
    if (jp == null) {
        HdfsUsers user = hdfsUsersFacade.findByName(hdfsUser);
        String configSecret = DigestUtils.sha256Hex(Integer.toString(ThreadLocalRandom.current().nextInt()));
        JupyterDTO dto = null;
        DistributedFileSystemOps dfso = dfsService.getDfsOps();
        String allowOriginHost = uriInfo.getBaseUri().getHost();
        int allowOriginPort = uriInfo.getBaseUri().getPort();
        String allowOriginPortStr = allowOriginPort != -1 ? ":" + allowOriginPort : "";
        String allowOrigin = settings.getJupyterOriginScheme() + "://" + allowOriginHost + allowOriginPortStr;
        try {
            jupyterSettingsFacade.update(jupyterSettings);
            // Inspect dependencies
            sparkController.inspectDependencies(project, hopsworksUser, (SparkJobConfiguration) jupyterSettings.getJobConfig());
            dto = jupyterManager.startJupyterServer(project, configSecret, hdfsUser, hopsworksUser, jupyterSettings, allowOrigin);
            jupyterJWTManager.materializeJWT(hopsworksUser, project, jupyterSettings, dto.getCid(), dto.getPort(), JUPYTER_JWT_AUD);
            HopsUtils.materializeCertificatesForUserCustomDir(project.getName(), user.getUsername(), settings.getHdfsTmpCertDir(), dfso, certificateMaterializer, settings, dto.getCertificatesDir());
            jupyterManager.waitForStartup(project, hopsworksUser);
        } catch (ServiceException | TimeoutException ex) {
            if (dto != null) {
                jupyterController.shutdownQuietly(project, hdfsUser, hopsworksUser, configSecret, dto.getCid(), dto.getPort());
            }
            throw new ServiceException(RESTCodes.ServiceErrorCode.JUPYTER_START_ERROR, Level.SEVERE, ex.getMessage(), null, ex);
        } catch (IOException ex) {
            if (dto != null) {
                jupyterController.shutdownQuietly(project, hdfsUser, hopsworksUser, configSecret, dto.getCid(), dto.getPort());
            }
            throw new HopsSecurityException(RESTCodes.SecurityErrorCode.CERT_MATERIALIZATION_ERROR, Level.SEVERE, ex.getMessage(), null, ex);
        } finally {
            if (dfso != null) {
                dfsService.closeDfsClient(dfso);
            }
        }
        String externalIp = Ip.getHost(req.getRequestURL().toString());
        try {
            Date expirationDate = new Date();
            Calendar cal = Calendar.getInstance();
            cal.setTime(expirationDate);
            cal.add(Calendar.HOUR_OF_DAY, jupyterSettings.getShutdownLevel());
            expirationDate = cal.getTime();
            jp = jupyterFacade.saveServer(externalIp, project, configSecret, dto.getPort(), user.getId(), dto.getToken(), dto.getCid(), expirationDate, jupyterSettings.isNoLimit());
            // set minutes left until notebook server is killed
            Duration durationLeft = Duration.between(new Date().toInstant(), jp.getExpires().toInstant());
            jp.setMinutesUntilExpiration(durationLeft.toMinutes());
        } catch (Exception e) {
            LOGGER.log(Level.SEVERE, "Failed to save Jupyter notebook settings", e);
            jupyterController.shutdownQuietly(project, hdfsUser, hopsworksUser, configSecret, dto.getCid(), dto.getPort());
        }
        if (jp == null) {
            throw new ServiceException(RESTCodes.ServiceErrorCode.JUPYTER_SAVE_SETTINGS_ERROR, Level.SEVERE);
        }
        if (jupyterSettings.isGitBackend()) {
            try {
                // Init is idempotent, calling it on an already initialized repo won't affect it
                jupyterNbVCSController.init(jp, jupyterSettings);
                if (jupyterSettings.getGitConfig().getStartupAutoPull()) {
                    jupyterNbVCSController.pull(jp, jupyterSettings);
                }
            } catch (ServiceException ex) {
                jupyterController.shutdownQuietly(project, hdfsUser, hopsworksUser, configSecret, dto.getCid(), dto.getPort());
                throw ex;
            }
        }
    } else {
        throw new ServiceException(RESTCodes.ServiceErrorCode.JUPYTER_SERVER_ALREADY_RUNNING, Level.FINE);
    }
    return noCacheResponse.getNoCacheResponseBuilder(Response.Status.OK).entity(jp).build();
}
Also used : DistributedFileSystemOps(io.hops.hopsworks.common.hdfs.DistributedFileSystemOps) Calendar(java.util.Calendar) JupyterProject(io.hops.hopsworks.persistence.entity.jupyter.JupyterProject) Duration(java.time.Duration) HdfsUsers(io.hops.hopsworks.persistence.entity.hdfs.user.HdfsUsers) Users(io.hops.hopsworks.persistence.entity.user.Users) IOException(java.io.IOException) HdfsUsers(io.hops.hopsworks.persistence.entity.hdfs.user.HdfsUsers) Date(java.util.Date) TimeoutException(java.util.concurrent.TimeoutException) ProjectException(io.hops.hopsworks.exceptions.ProjectException) JobException(io.hops.hopsworks.exceptions.JobException) GenericException(io.hops.hopsworks.exceptions.GenericException) HopsSecurityException(io.hops.hopsworks.exceptions.HopsSecurityException) ElasticException(io.hops.hopsworks.exceptions.ElasticException) IOException(java.io.IOException) ServiceException(io.hops.hopsworks.exceptions.ServiceException) HopsSecurityException(io.hops.hopsworks.exceptions.HopsSecurityException) ProjectException(io.hops.hopsworks.exceptions.ProjectException) ServiceException(io.hops.hopsworks.exceptions.ServiceException) GitConfig(io.hops.hopsworks.persistence.entity.jupyter.config.GitConfig) YarnProjectsQuota(io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota) JupyterDTO(io.hops.hopsworks.common.dao.jupyter.config.JupyterDTO) TimeoutException(java.util.concurrent.TimeoutException) Path(javax.ws.rs.Path) POST(javax.ws.rs.POST) Consumes(javax.ws.rs.Consumes) Produces(javax.ws.rs.Produces) JWTRequired(io.hops.hopsworks.jwt.annotation.JWTRequired) AllowedProjectRoles(io.hops.hopsworks.api.filter.AllowedProjectRoles)

Example 2 with YarnProjectsQuota

use of io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota in project hopsworks by logicalclocks.

the class ProjectController method removeQuotas.

@TransactionAttribute(TransactionAttributeType.REQUIRES_NEW)
private void removeQuotas(Project project) {
    YarnProjectsQuota yarnProjectsQuota = yarnProjectsQuotaFacade.findByProjectName(project.getName());
    yarnProjectsQuotaFacade.remove(yarnProjectsQuota);
}
Also used : YarnProjectsQuota(io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota) TransactionAttribute(javax.ejb.TransactionAttribute)

Example 3 with YarnProjectsQuota

use of io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota in project hopsworks by logicalclocks.

the class ProjectController method setProjectOwnerAndQuotas.

public void setProjectOwnerAndQuotas(Project project, DistributedFileSystemOps dfso, Users user) throws IOException {
    this.yarnProjectsQuotaFacade.persistYarnProjectsQuota(new YarnProjectsQuota(project.getName(), settings.getYarnDefaultQuota(), 0));
    this.yarnProjectsQuotaFacade.flushEm();
    // Here we set only the project quota. The HiveDB and Feature Store quotas are set in the HiveController
    if (settings.getHdfsDefaultQuotaInMBs() > -1) {
        dfso.setHdfsSpaceQuotaInMBs(new Path(Utils.getProjectPath(project.getName())), settings.getHdfsDefaultQuotaInMBs());
    }
    projectFacade.setTimestampQuotaUpdate(project, new Date());
    // Add the activity information
    logActivity(ActivityFacade.NEW_PROJECT + project.getName(), user, project, ActivityFlag.PROJECT);
    // update role information in project
    addProjectOwner(project, user);
    LOGGER.log(Level.FINE, "{0} - project created successfully.", project.getName());
}
Also used : Path(org.apache.hadoop.fs.Path) Date(java.util.Date) YarnProjectsQuota(io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota)

Example 4 with YarnProjectsQuota

use of io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota in project hopsworks by logicalclocks.

the class YarnProjectsQuotaFacade method changeYarnQuota.

public void changeYarnQuota(String projectName, float quota) {
    YarnProjectsQuota project = findByProjectName(projectName);
    if (project != null) {
        project.setQuotaRemaining(quota);
        em.merge(project);
    }
}
Also used : YarnProjectsQuota(io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota)

Example 5 with YarnProjectsQuota

use of io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota in project hopsworks by logicalclocks.

the class AbstractExecutionController method start.

@Override
@TransactionAttribute(TransactionAttributeType.NOT_SUPPORTED)
public Execution start(Jobs job, String args, Users user) throws JobException, GenericException, ServiceException, ProjectException {
    // If the limit for the number of executions for this job has been reached, return an error
    checkExecutionLimit(job);
    // A user should not be able to start a job if the project is prepaid and it doesn't have quota.
    if (job.getProject().getPaymentType().equals(PaymentType.PREPAID)) {
        YarnProjectsQuota projectQuota = yarnProjectsQuotaFacade.findByProjectName(job.getProject().getName());
        if (projectQuota == null || projectQuota.getQuotaRemaining() <= 0) {
            throw new ProjectException(RESTCodes.ProjectErrorCode.PROJECT_QUOTA_ERROR, Level.FINE);
        }
    }
    // If enabled and nodemanagers are all offline throw an JobException exception
    if (settings.isCheckingForNodemanagerStatusEnabled() && job.getJobType() != JobType.PYTHON) {
        hostServicesFacade.findServices("nodemanager").stream().filter(s -> s.getStatus() == ServiceStatus.Started).findFirst().orElseThrow(() -> new JobException(RESTCodes.JobErrorCode.NODEMANAGERS_OFFLINE, Level.SEVERE));
    }
    Execution exec;
    switch(job.getJobType()) {
        case FLINK:
            // Materialize certs
            return flinkController.startJob(job, user);
        case SPARK:
            exec = sparkController.startJob(job, args, user);
            if (exec == null) {
                throw new IllegalArgumentException("Problem getting execution object for: " + job.getJobType());
            }
            SparkJobConfiguration config = (SparkJobConfiguration) job.getJobConfig();
            String path = config.getAppPath();
            String pathOfInode;
            try {
                pathOfInode = Utils.prepPath(path);
            } catch (UnsupportedEncodingException ex) {
                throw new JobException(RESTCodes.JobErrorCode.JOB_START_FAILED, Level.FINE, "Job name: " + job.getName(), ex.getMessage(), ex);
            }
            Inode inode = inodeController.getInodeAtPath(pathOfInode);
            String inodeName = inode.getInodePK().getName();
            activityFacade.persistActivity(ActivityFacade.EXECUTED_JOB + inodeName, job.getProject(), user, ActivityFlag.JOB);
            break;
        case PYSPARK:
            if (job.getProject().getPythonEnvironment() == null) {
                throw new ProjectException(RESTCodes.ProjectErrorCode.ANACONDA_NOT_ENABLED, Level.FINEST);
            }
            exec = sparkController.startJob(job, args, user);
            if (exec == null) {
                throw new IllegalArgumentException("Error while getting execution object for: " + job.getJobType());
            }
            break;
        default:
            throw new GenericException(RESTCodes.GenericErrorCode.UNKNOWN_ACTION, Level.FINE, "Unsupported job type: " + job.getJobType());
    }
    return exec;
}
Also used : ProjectException(io.hops.hopsworks.exceptions.ProjectException) JobException(io.hops.hopsworks.exceptions.JobException) Execution(io.hops.hopsworks.persistence.entity.jobs.history.Execution) Inode(io.hops.hopsworks.persistence.entity.hdfs.inode.Inode) SparkJobConfiguration(io.hops.hopsworks.persistence.entity.jobs.configuration.spark.SparkJobConfiguration) UnsupportedEncodingException(java.io.UnsupportedEncodingException) GenericException(io.hops.hopsworks.exceptions.GenericException) YarnProjectsQuota(io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota) TransactionAttribute(javax.ejb.TransactionAttribute)

Aggregations

YarnProjectsQuota (io.hops.hopsworks.persistence.entity.jobs.quota.YarnProjectsQuota)6 GenericException (io.hops.hopsworks.exceptions.GenericException)2 JobException (io.hops.hopsworks.exceptions.JobException)2 ProjectException (io.hops.hopsworks.exceptions.ProjectException)2 Date (java.util.Date)2 TransactionAttribute (javax.ejb.TransactionAttribute)2 AllowedProjectRoles (io.hops.hopsworks.api.filter.AllowedProjectRoles)1 JupyterDTO (io.hops.hopsworks.common.dao.jupyter.config.JupyterDTO)1 DistributedFileSystemOps (io.hops.hopsworks.common.hdfs.DistributedFileSystemOps)1 ElasticException (io.hops.hopsworks.exceptions.ElasticException)1 HopsSecurityException (io.hops.hopsworks.exceptions.HopsSecurityException)1 ServiceException (io.hops.hopsworks.exceptions.ServiceException)1 JWTRequired (io.hops.hopsworks.jwt.annotation.JWTRequired)1 Dataset (io.hops.hopsworks.persistence.entity.dataset.Dataset)1 HdfsDirectoryWithQuotaFeature (io.hops.hopsworks.persistence.entity.hdfs.HdfsDirectoryWithQuotaFeature)1 Inode (io.hops.hopsworks.persistence.entity.hdfs.inode.Inode)1 HdfsUsers (io.hops.hopsworks.persistence.entity.hdfs.user.HdfsUsers)1 SparkJobConfiguration (io.hops.hopsworks.persistence.entity.jobs.configuration.spark.SparkJobConfiguration)1 Execution (io.hops.hopsworks.persistence.entity.jobs.history.Execution)1 JupyterProject (io.hops.hopsworks.persistence.entity.jupyter.JupyterProject)1