Line: 1 to 1 | ||||||||
---|---|---|---|---|---|---|---|---|
CREAM User's Guide for EMI-3 | ||||||||
Line: 896 to 896 | ||||||||
| ||||||||
Deleted: | ||||||||
< < | ||||||||
Added: | ||||||||
> > |
|
Line: 1 to 1 | ||||||||
---|---|---|---|---|---|---|---|---|
CREAM User's Guide for EMI-3 | ||||||||
Line: 895 to 895 | ||||||||
| ||||||||
Added: | ||||||||
> > |
Line: 1 to 1 | ||||||||
---|---|---|---|---|---|---|---|---|
CREAM User's Guide for EMI-3 | ||||||||
Line: 80 to 80 | ||||||||
| ||||||||
Added: | ||||||||
> > |
| |||||||
Man pages are available for all the CREAM client commands. You can also access information about the usage of each command by issuing: | ||||||||
Line: 117 to 118 | ||||||||
glite-ce-allowed-submission checks if jobs submissions on the specified CREAM CE are allowed or have been disabled. | ||||||||
Added: | ||||||||
> > | glite-ce-job-lease create a lease identifier in the CREAM server and associate a time duration to it. | |||||||
All these commands are described in the following sections.
Submitting jobs to CREAM based CEs | ||||||||
Line: 171 to 174 | ||||||||
The command returns the CREAM job identifiers associated with these jobs (e.g. https://cream-ce-01.pd.infn.it:8443/CREAM116j9vgnf ) which identify them in clear and unique way all over the Grid system scope. | ||||||||
Added: | ||||||||
> > | In addition the user can associate a lease that she/he has previously created with the command glite-ce-job-lease by mean of the option --leaseId :
> glite-ce-job-submit -D mydelid -r cream-ce-01.pd.infn.it:8443/cream-lsf-grid02 --leaseId <my_lease_identifier> myjob1.jdl myjob2.jdl myjob3.jdlTo create a lease in the CREAM service, with a certain duration of time (expressed in seconds), issue the command: glite-ce-job-lease --endpoint cream-27.pd.infn.it --leaseTime 3600 myLID You requested lease time [3600] for lease ID [myLID] CREAM negotiated the lease time to [3600]The above command has created a lease on cream-27.pd.infn.it named "myLID" and lasting 1 hour. | |||||||
Monitoring jobsPassing the CREAM job identifiers returned by theglite-ce-job-submit command to the glite-ce-job-status command, it is possible to monitor the submitted jobs. Several (static and dynamic) information can be shown, depending on the chosen verbosity level. The verbosity level can be 0 (less verbosity), 1 or 2 (most verbosity). | ||||||||
Line: 461 to 480 | ||||||||
| ||||||||
Added: | ||||||||
> > |
| |||||||
As mentioned above, if the same attribute is defined in more than a configuration file, the definition in the user specific configuration file (if any) has higher priority than the definition in the VO specific configuration file (if any), which has higher priority than the definition in the generic configuration file. If an attribute is not defined anywhere, the default value is considered. | ||||||||
Line: 483 to 505 | ||||||||
SUSPEND_LOG_DIR="tmp/CREAMLogs" LIST_LOG_DIR="tmp/CREAMLogs" DELEGATE_LOG_DIR="tmp/CREAMLogs" | ||||||||
Added: | ||||||||
> > | LEASE_LOG_DIR="tmp/CREAMLogs" | |||||||
] | ||||||||
Line: 504 to 527 | ||||||||
Added: | ||||||||
> > | ||||||||
Use specific functionality of the CREAM CE |
Line: 1 to 1 | ||||||||
---|---|---|---|---|---|---|---|---|
Added: | ||||||||
> > | CREAM User's Guide for EMI-3
CREAM Command Line Interface GuideThis section briefly explains the sequence of operations to be performed by a user to submit and then manage jobs on CREAM based CEs, referring to the C++ Command Line Interface (CLI).Before starting: get your user proxyBefore using any of the CREAM client commands, it is necessary to have a valid proxy credential available on the client machine. You can create it using thevoms-proxy-init command. If you already have a valid proxy available on your machine just make the X509_USER_PROXY environment variable point to it.
In order to get a proxy certificate issued by VOMS, you should have in the directory /etc/vomses the proper VOMS file containing a line as follows:
"EGEE" "kuiken.nikhef.nl" "15001" "/O=dutchgrid/O=hosts/OU=nikhef.nl/CN=kuiken.nikhef.nl" "EGEE" "22"or the corresponding line for your VO. You also need to install the VO related .lsc files in the /etc/grid-security/vomsdir/<VO> directory. In a standard EMI UI installation, these settings should be already there.
Make moreover sure you have in the directory $HOME/.globus your certificate/key pair, i.e. the following files:
usercert.pem userkey.pemNote that file permissions are important: the two files must have respectively 0600 and 0400 permissions. Then you can issue the VOMS client command (you will be prompted for the pass-phrase): $ voms-proxy-init -voms dteam Enter GRID pass phrase: Your identity: /C=IT/O=INFN/OU=Personal Certificate/L=Padova/CN=Massimo Sgaravatto Creating temporary proxy ............................................................................................................................................... Done Contacting voms2.hellasgrid.gr:15004 [/C=GR/O=HellasGrid/OU=hellasgrid.gr/CN=voms2.hellasgrid.gr] "dteam" Done Creating proxy .............................. Done Your proxy is valid until Sat Apr 30 05:05:49 2011 $ voms-proxy-info -all subject : /C=IT/O=INFN/OU=Personal Certificate/L=Padova/CN=Massimo Sgaravatto/CN=proxy issuer : /C=IT/O=INFN/OU=Personal Certificate/L=Padova/CN=Massimo Sgaravatto identity : /C=IT/O=INFN/OU=Personal Certificate/L=Padova/CN=Massimo Sgaravatto type : proxy strength : 1024 bits path : /tmp/x509up_u500 timeleft : 11:59:55 key usage : Digital Signature, Key Encipherment, Data Encipherment === VO dteam extension information === VO : dteam subject : /C=IT/O=INFN/OU=Personal Certificate/L=Padova/CN=Massimo Sgaravatto issuer : /C=GR/O=HellasGrid/OU=hellasgrid.gr/CN=voms2.hellasgrid.gr attribute : /dteam/Role=NULL/Capability=NULL attribute : /dteam/italy/Role=NULL/Capability=NULL attribute : /dteam/italy/INFN-PADOVA/Role=NULL/Capability=NULL timeleft : 11:59:55 uri : voms2.hellasgrid.gr:15004 CREAM CLI commandsThe most relevant commands to interact with CREAM based CEs are:
$ <command> --help glite-ce-job-submit submits N jobs (N must be >=1) to a CREAM based CE. It requires N JDL files as input and returns N CREAM job identifiers.
glite-ce-delegate-proxy allows the user to delegate her proxy credential to the CREAM service. This delegated credential can then be used for job submissions.
glite-ce-job-status displays information (in particular the states) of N jobs (N must be >=1) previously submitted to CREAM based CEs.
glite-ce-job-list lists the identifiers of jobs submitted to a CREAM based CE by the user issuing the command.
glite-ce-job-cancel cancels N jobs (N must be >=1) previously submitted to CREAM based CEs.
glite-ce-job-suspend suspends the execution of N jobs (N must be >=1) previously submitted to CREAM based CEs.
glite-ce-job-resume resumes the execution of N jobs (N must be >=1) which have been previously suspended.
glite-ce-job-output retrieves the output sandbox files of N jobs (N must be >=1) previously submitted to CREAM based CEs.
glite-ce-job-purge clears N jobs (N must be >=1) from CREAM based CEs. After this operation the purged jobs can’t be managed anymore.
glite-ce-proxy-renew renews N delegations (N must be >=1), and therefore refreshes the proxy of the jobs submitted to CREAM based CEs using the considered delegations.
glite-ce-service-info returns information about the CREAM service (version, status, etc.).
glite-ce-get-cemon-url returns the end-point of the CEMon service coupled with the considered CREAM CE.
glite-ce-enable-submission (re-)enables job submissions on the specified CREAM CE.
glite-ce-disable-submission disables job submissions on the specified CREAM CE.
glite-ce-allowed-submission checks if jobs submissions on the specified CREAM CE are allowed or have been disabled.
All these commands are described in the following sections.
Submitting jobs to CREAM based CEsTo submit jobs to CREAM based CEs, the commandglite-ce-job-submit must be used. The glite-ce-job-submit command requires as input one or more job description files; each file describes the job characteristics and requirements through the JDL (Job Description Language). A typical example of a JDL job description file is:
[ Type = "Job"; JobType = "Normal"; Executable = "myexe"; StdInput = "myinput.txt"; StdOutput = "message.txt"; StdError = "error.txt"; InputSandbox = {"/users/seredova/example/myinput.txt", "/users/seredova/example/myexe"}; OutputSandbox = {"message.txt", "error.txt"}; OutputSandboxBaseDestUri = "gsiftp://se.pd.infn.it/data/seredova"; ]Such a JDL would make the myexe executable be transferred on the remote CREAM CE and be run taking the myinput.txt file (also copied from the client node) as input. The standard streams of the job are redirected to files message.txt and error.txt , and when job completes its execution they are automatically uploaded on gsiftp://se.pd.infn.it/data/seredova .
A detailed description of the available JDL attributes and of the rules for building correct JDL files is provided at http://wiki.italiangrid.org/twiki/bin/view/CREAM/JdlGuide![]()
glite-ce-delegate-proxy is the command to be used to explicitly delegate the user credentials to a CREAM CE.
The following shows an example of job submission, performed explicitly delegating credentials. So first of all the credentials are delegated to a CREAM based CE (whose endpoint is specified with the option --endpoint (-e ):
> glite-ce-delegate-proxy -e cream-ce-01.pd.infn.it mydelid 2006-02-26 15:03:37,286 NOTICE - Proxy with delegation id [mydelid] successfully delegated to endpoint [https://cream-ce-01.pd.infn.it:8443//ce-cream/services/CREAMDelegation]The identifier of the delegation is then specified with the --delegationId (-D ) option in the job submit operation:
> glite-ce-job-submit -D mydelid -r cream-ce-01.pd.infn.it:8443/cream-lsf-grid02 myjob1.jdl myjob2.jdl myjob3.jdlThe option -r (--resource ) has been used to specify the identifier of the CREAM CE where the job has to be submitted to.
myjob1.jdl myjob2.jdl myjob3.jdl are the 3 JDL files describing the jobs to be submitted.
The command returns the CREAM job identifiers associated with these jobs (e.g. https://cream-ce-01.pd.infn.it:8443/CREAM116j9vgnf ) which identify them in clear and unique way all over the Grid system scope.
Monitoring jobsPassing the CREAM job identifiers returned by theglite-ce-job-submit command to the glite-ce-job-status command, it is possible to monitor the submitted jobs. Several (static and dynamic) information can be shown, depending on the chosen verbosity level. The verbosity level can be 0 (less verbosity), 1 or 2 (most verbosity).
Please note that specifying 0 as verbosity level means calling on the CREAM service a faster operation than when using 1 or 2 as verbosity level.
The most relevant attribute is the job status.
The following is an example of job status operation, specifying 1 as verbosity level:
$ glite-ce-job-status -L 1 https://cream-02.pd.infn.it:8443/CREAM738582717 ****** JobID=[https://cream-02.pd.infn.it:8443/CREAM738582717] Current Status = [DONE-FAILED] ExitCode = [N/A] FailureReason = [lsf_reason=256; Cannot move ISB (${globus_transfer_cmd} gsiftp://cream-02.pd.infn.it//CREAMTests/Exe1/ssh1.sh file:///home/infngrid001/home_cream_738582717/CREAM738582717/ssh1.sh): error: globus_ftp_client: the server responded with an error 500 500-Command failed. : globus_l_gfs_file_open failed. 500-globus_xio: Unable to open file //CREAMTests/Exe1/ssh1.sh 500-globus_xio: System error in open: No such file or directory 500-globus_xio: A system call failed: No such file or directory 500 End.] Grid JobID = [N/A] Job status changes: ------------------- Status = [REGISTERED] - [Tue 22 Jan 2008 15:55:08] (1201013708) Status = [PENDING] - [Tue 22 Jan 2008 15:55:08] (1201013708) Status = [IDLE] - [Tue 22 Jan 2008 15:55:11] (1201013711) Status = [RUNNING] - [Tue 22 Jan 2008 15:55:18] (1201013718) Status = [DONE-FAILED] - [Tue 22 Jan 2008 16:03:10] (1201014190) Issued Commands: ------------------- *** Command Name = [JOB_REGISTER] Command Category = [JOB_MANAGEMENT] Command Status = [SUCCESSFULL] *** Command Name = [JOB_START] Command Category = [JOB_MANAGEMENT] Command Status = [SUCCESSFULL]In this example it is interesting to note that the job failed (as reported by the Current Status field) for the problem reported in the FailureReason field: the file to be transferred was not found.
Instead of explicitly specifying the identifiers of the jobs to monitor, the user can also ask to monitor all her jobs, in case specifying conditions (on the submission date and/or on the job status) that must be met.
For example to monitor all jobs, whose status is DONE-OK or DONE-FAILED, submitted to the grid005.pd.infn.it CREAM CE between July 23, 2005 10:00 and July 28, 2005 11:00, the following command must be issued:
> glite-ce-job-status --all -e grid005.pd.infn.it:8443 --from ’2005-07-23 10:00:00’ --to ’2005-07-28 11:00:00’ -s DONE-OK:DONE-FAILED Retrieving output of jobsUser can choose to save the output sandbox (OSB) files on a remote server, or save them in the CREAM CE node. In the latter case these files can then be retrieved using theglite-ce-job-output command.
For example the following command retrieves the output sandbox files of the specified job from the relevant CREAM CE node:
> glite-ce-job-output https://cream-38.pd.infn.it:8443/CREAM295728364 2011-01-29 10:09:50,394 INFO - For JobID [https://cream-38.pd.infn.it:8443/CREAM295728364] output will be stored in the dir ./cream-38.pd.infn.it_8443_CREAM295728364This command can be used also to retrieve output produced by multiple jobs, by specifying multiple job identifiers as command's arguments . Getting job identifiersIf a user is interested to get the identifiers of all her jobs submitted to a specific CREAM CE, she can use theglite-ce-job-list command. For example the following command returns the identifiers of all the jobs submitted to the specified CREAM CE, owned by the user issuing the command:
> glite-ce-job-list grid005.pd.infn.it:8443 Cancelling jobsIn some cases it might be needed to cancel jobs which have been previously submitted to CREAM based CEs. This can be achieved via theglite-ce-job-cancel command.
E.g., the command:
> glite-ce-job-cancel https://grid005.pd.infn.it:8443/CREAM115j5vfnfcancels the specified job. Suspending and resuming jobsA running or idle job can be suspended (i.e. its execution will be stopped), and be resumed (i.e. it will run again) later. This can be achieved with theglite-ce-job-suspend and glite-ce-job-resume commands.
The following example shows that after having issued the glite-ce-job-suspend command, after a while the job status becomes HELD .
> glite-ce-job-suspend https://cream-ce-01.pd.infn.it:8443/CREAM11a79tnb2 Are you sure you want to suspend specified job(s) [y/n]: y > glite-ce-job-status -L 0 https://cream-ce-01.pd.infn.it:8443/CREAM11a79tnb2 ****** JobID=[https://cream-ce-01.pd.infn.it:8443/CREAM11a79tnb2] Status = [HELD]Issuing the glite-ce-job-resume command, the job will run/will be idle again:
> glite-ce-job-resume https://cream-ce-01.pd.infn.it:8443/CREAM11a79tnb2 Are you sure you want to resume specified job(s) [y/n]: y > glite-ce-job-status -L 0 https://cream-ce-01.pd.infn.it:8443/CREAM11a79tnb2 ****** JobID=[https://cream-ce-01.pd.infn.it:8443/CREAM11a79tnb2] Status = [REALLY-RUNNING] Purging jobsA CREAM job can be monitored (via theglite-ce-job-status ) even after it has completed its execution. A job gets “lost” (i.e. it is not possible to monitor or manage it anymore) only when the user who submitted it decides to explicitly clear it, or when the CREAM system administrator decides to do this purging operation. A user can purge her own jobs, using the glite-ce-job-purge command.
E.g., after having issued the command:
> glite-ce-job-purge https://cream-ce-01.pd.infn.it:8443/CREAM116jbi4o0the specified job can’t be managed anymore (e.g. it is not possible to check its status anymore). Renewing proxiesIt is possible that long jobs may outlive the validity of the initial delegated credentials; if so the job will die prematurely. To avoid this it is possible to renew the proxy of jobs submitted to CREAM CEs with theglite-ce-proxy-renew command.
E.g. the following command:
> glite-ce-proxy-renew -e cream-ce-01.pd.infn.it:8443 mydelidrenews the proxy of all the jobs having mydelid as delegation id.
It must be stressed that for jobs submitted to CREAM based CEs via the Workload Management System (WMS), proxy renewal is automatically dealt by the middleware.
Handling job identifiersHandling the job identifiers directly quickly becomes tedious. To avoid this, you can make theglite-ce-job-submit and glite-ce-job-list commands append the job Id(s) to a named file using the --output (-o ) option. On the other side, the CREAM client commands which take job identifier(s) as argument accept also the --input (-i ) option which allows the job identifier(s) to be read from a file.
The following shows an example:
> glite-ce-job-submit -a -r cream-ce-01.pd.infn.it:8443/cream-lsf-grid02 -o idfile myjob.jdl https://cream-ce-01.pd.infn.it:8443/CREAM116jbs5b9The returned job id got also inserted in the specified file ( idfile ), which can be specified with the --input (-i ) option e.g. with the glite-ce-job-status command:
> glite-ce-job-status -i idfile ****** JobID=[https://cream-ce-01.pd.infn.it:8443/CREAM116jbs5b9] Status=[REALLY-RUNNING] Restricting job submissionsIn order to prevent that a CREAM CE gets overloaded, the CREAM CE administrator can set a specific policy to disable new job submissions when certain conditions are met. If submissions are disabled because of that, if newer job submissions are attempted, users will get an error message such as:> glite-ce-job-submit -a -r cream-38.pd.infn.it:8443/cream-pbs-creamtest1 oo.jdl MethodName=[jobRegister] ErrorCode=[0] Description=[The CREAM service cannot accept jobs at the moment] FaultCause=[Threshold for Load Average(1 min): 30 => Detected value for Load Average(1 min): 31.13] Timestamp=[Sat 29 Jan 2011 11:55:18]In order to avoid degrading the performance of the system, the specified policy is not evaluated for each job submission, but instead it is evaluated and imposed from time to time (so it might happen that for a short time job submissions are allowed even if the specified threshold has been reached). CREAM “super-users” can also disable newer job submissions via the command glite-ce-disable-submission . Submissions can then be re-enabled by a CREAM “super-user” via the command glite-ce-enable-submission .
To check if job submissions on a specific CREAM CE are allowed, the command glite-ce-allowed-submission can be used.
E.g.:
> glite-ce-disable-submission grid006.pd.infn.it:8443 Operation for disabling new submissions succeeded > > glite-ce-allowed-submission grid006.pd.infn.it:8443 Job Submission to this CREAM CE is disabled > > glite-ce-enable-submission grid006.pd.infn.it:8443 Operation for enabling new submissions succeeded > > glite-ce-allowed-submission grid006.pd.infn.it:8443 Job Submission to this CREAM CE is enabledIt must be stressed that if job submissions to a specific CREAM CE are disabled, all other operations (job status, job cancellations, etc.) can still be performed. Getting information about the CREAM serviceIt is possible to get information about the CREAM service (interface and service version, status, etc) using theglite-ce-service-info command, e.g.:
> glite-ce-service-info cream-13.pd.infn.it:8443 Interface Version = [2.1] Service Version = [1.12] Description = [CREAM 2] Started at = [Tue Nov 10 14:42:12 2009] Submission enabled = [YES] Status = [RUNNING] Service Property = [SUBMISSION_THRESHOLD_MESSAGE]-> [Threshold for Load Average (1 min): 10 => Detected value for Load Average(1 min): 0.03 Threshold for Load Average(5 min): 10 => Detected value for Load Average(5 min): 0.03 Threshold for Load Average(15 min): 10 => Detected value for Load Average(15 min): 0.00 Threshold for Memory Usage: 95 => Detected value for Memory Usage: 57.41% Threshold for Swap Usage: 95 => Detected value for Swap Usage: 2.02% Threshold for Free FD: 500 => Detected value for Free FD: 204500 Threshold for tomcat FD: 800 => Detected value for Tomcat FD: 107 Threshold for FTP Connection: 30 => Detected value for FTP Connection: 1 Threshold for Number of active jobs: -1 => Detected value for Number of active jobs: 0 Threshold for Number of pending commands: -1 => Detected value for Number of pending commands: 0A CREAM CE is usually coupled with a CEMon service, which can be queried to get information about the CE and/or can notify clients with specific CE events. The command glite-ce-get-cemon-url can be used to get the end-point of this CEMon service, e.g.:
> glite-ce-get-cemon-url grid005.pd.infn.it:8443 https://grid005.pd.infn.it:8443/ce-monitor/services/CEMonitor CREAM CLI configuration filesThe configuration of the CREAM UI is accomplished via three possible configuration files:
CREAM CLI configuration file attributesWe list here the possible attributes that can be specified in the configuration files:
Example of CREAM CLI configuration fileThe following represents an example of a CREAM UI configuration file:[ JDL_DEFAULT_ATTRIBUTES = [ JobType=" Normal" ; Type="job" ]; STATUS_VERBOSITY_LEVEL = 2; CANCEL_LOG_DIR="tmp/CREAMLogs" PURGE_LOG_DIR="tmp/CREAMLogs" RESUME_LOG_DIR="tmp/CREAMLogs" STATUS_LOG_DIR="tmp/CREAMLogs" SUBMIT_LOG_DIR="tmp/CREAMLogs" SUSPEND_LOG_DIR="tmp/CREAMLogs" LIST_LOG_DIR="tmp/CREAMLogs" DELEGATE_LOG_DIR="tmp/CREAMLogs" ] Man pages for CREAM Command Line Interface
Use specific functionality of the CREAM CESubmission on multi-core resourcesAs explained in the CREAM JDL guide![]()
First scenarioWith a JDL such as:WholeNodes=true; SMPGranularity=G; Hostnumber=H;with H > 1. In the submission script there will be:
BSUB -n S*H BSUB -R "span[ptile=S] BSUB -x
PBS -l nodes=H:ppn=S PBS -W x=NACCESSPOLICY:SINGLEJOBwith S equal to the value published as GlueHostArchitectureSMPSize .
Second scenarioWith a JDL such as:WholeNodes=true; SMPGranularity=G;in the submission script there will be:
BSUB -n S BSUB -R "span[hosts=1]" BSUB -x
PBS -l nodes=1:ppn=S PBS -W x=NACCESSPOLICY:SINGLEJOB Third scenarioWith a JDL such as:WholeNodes=true; HostNumber=H;with H>1. in the submission script there will be:
BSUB -n S*H BSUB -R "span[ptile=S]" BSUB -x
PBS -l nodes=H:ppn=S PBS -W x=NACCESSPOLICY:SINGLEJOBwith S equal to the value published as GlueHostArchitectureSMPSize .
Forth scenarioWith a JDL such as:WholeNodes=false; SMPGranularity=G; CPUNumber=C;in the submission script there will be:
BSUB -n C BSUB -R "span[ptile=G]"
PBS -l nodes=N:ppn=G { [+1:ppn=R] if r>0 }with: N = C / G R = C % G Fifth scenarioWith a JDL such as:WholeNodes=false; HostNumber=H; CPUNumber=C;with H>=1. in the submission script there will be:
BSUB -n C BSUB -R "span[ptile={ N if R=0 ; N+1 if R>0 }]"
PBS -l nodes=H-R:ppn=N { [+R:ppn=N+1] if R>0 }with: N = C / H R = C % H Sixth scenarioWith a JDL such as:WholeNodes=false; CPUNumber=C;in the submission script there will be:
BSUB -n C
PBS -l nodes=C Forward of requirements to the batch systemThe CREAM CE allows to forward, via tha BLAH component, requirements to the batch system. For this purpose the JDLCERequirements attribute, described at http://wiki.italiangrid.org/twiki/bin/view/CREAM/JdlGuide#3_27_CERequirements![]() glite-ce-job-submit command) the CeRequirements attribute is supposed to be filled by the end-user.
For jobs submitted to the CREAM CE via the WMS, the CeRequirements attribute is instead filled by the WMS, considering the JDL Requirements expression and the value of the CeForwardParameters attribute in the WMS configuration file.
For example, if in the user JDL there is :
Requirements= "other.GlueHostMainMemoryRAMSize > 100 && other.GlueCEImplementationName==\"CREAM\"";and if the WMS configuration file there is: CeForwardParameters = {"GlueHostMainMemoryVirtualSize","GlueHostMainMemoryRAMSize", "GlueCEPolicyMaxCPUTime"};in the JDL sent by the WMS to CREAM there will be: CeRequirements= "other.GlueHostMainMemoryRAMSize > 100";The CERequirements expression received by CREAM is then forwarded to BLAH. Basically BLAH manages the CERequirements expression setting some environment variables, which are available and can be properly used by the /usr/libexec/xxx_local_submit_attributes.sh script (e.g. /usr/libexec/pbs_local_submit_attributes.sh for PBS/Torque, /usr/libexec/lsf_local_submit_attributes.sh for LSF). This script must be properly created by the site admin.
For example, considering the following CeRequirements expression:
CeRequirements="other.GlueHostMainMemoryRAMSize > 100 && other.GlueCEStateWaitingJobs <10 && other.GlueCEImplementationName==\"CREAM\" && other.GlueHostProcessorClockSpeed >= 2800 && (Member(\"FDTD\", other.GlueHostApplicationSoftwareRuntimeEnvironment))";the following settings will be available in $USR_LOCATION/libexec/xxx_local_submit_attributes.sh :
GlueHostMainMemoryRAMSize_Min='100' GlueCEStateWaitingJobs_Max='10' GlueCEImplementationName='CREAM' GlueHostProcessorClockSpeed_Min='2800' GlueHostApplicationSoftwareRuntimeEnvironment='"FDTD"'where the value for $USR_LOCATION in a standard installation of a CREAM CE is "/usr". What is printed by the /usr/libexec/xxx_local_submit_attributes.sh script is automatically added to the submit command file.
For example if the JDL CeRequirements expression is:
CeRequirements = "(Member(\"FDTD\", other.GlueHostApplicationSoftwareRuntimeEnvironment))";and the /usr/libexec/pbs_local_submit_attributes.sh is:
#!/bin/sh if [ "$other.GlueHostApplicationSoftwareRuntimeEnvironment" == "FDTD" ]; then echo "#PBS -l software=FDTD" fithen the PBS submit file that will be used will include: ... ... # PBS directives: #PBS -S /bin/bash #PBS -o /dev/null #PBS -e /dev/null #PBS -l software=FDTD .... ....where the line: #PBS -l software=FDTDis set via the /usr/libexec/pbs_local_submit_attributes.sh script.
Please note that there are no differences if in CeRequirements expresssion there is e.g.
CeRequirements = other.xyz==\"ABC\"or: CeRequirements = "xyz==\"ABC\"";In both cases in /usr/libexec/xxx_local_submit_attributes.sh the variable xyz will be set.
As shown above, having x>a or x>=a doesn't make any difference in the setting of the environment variable x in the /usr/libexec/xxx_local_submit_attributes.sh script. It will be in both cases:
x_Min='a'Starting with EMI-2 (i.e. BLAH v. >= 1.18) it is possible to forward to the batch system also other attributes not included in the CeRequiments JDL attribute.
This can be done adding in /etc/blah.config the line:
blah_pass_all_submit_attributes=yesIn this way the xxx_local_submit_attributes.sh will see the following environment variables set:
blah.config e.g.:
blah_pass_submit_attributes[0]="x509UserProxySubject" blah_pass_submit_attributes[1]="x509UserProxyFQAN" CREAM job statesHere below is provided a brief description of the meaning of each possible state a CREAM job can enter:
![]() ES Command Line Interface Guide (WORK IN PROGRES...)Before starting: get your user proxySee above: get your user proxy![]() ES CLI Commands:
<hostname>[:TCPPORT] is a "pure" host's address optionally followed by ":" char and a number, and must not contain any leading protocol or trailing path.
GOOD endpoints are for example:
lxgrid05.pd.infn.it lxgrid05.pd.infn.it:443 lxgrid05.pd.infn.it:8080 devel08.cnaf.infn.it:2845When [:TCPPORT] is not specified, the default 8443 is automatically added.
BAD endpoints are:
http://devel15.cnaf.infn.it/ devel15.cnaf.infn.it/ devel15.cnaf.infn.it/somepath http://devel15.cnaf.infn.it:8443/anything https://devel15.cnaf.infn.it myprotocol://lxgrid05.pd.infn.itImmediate help can be printed on the console by issuing <command> --help glite-es-activity-create creates activities on the remote ES server specified as endpoint; multiple activities can be specified together in the same ADL_FILE (mandatory argument)
glite-es-activity-status obtains from the server, specified as endpoint, the status information about one or more activities; can take activities to query and endpoint from an input file
glite-es-activity-info obtains from the server, specified as endpoint, extended information about one or more activities; can take activities to query and endpoint from an input file
glite-es-activity-list obtains all the activity identifiers that have been created on the server specified as (mandatory) argument
glite-es-activty-pause tells the server, specified as endpoint, to pause the activity identifiers; can take activities to pause and endpoint from an input file
glite-es-activty-resume tells the server, specified as endpoint, to resume the activity identifiers; can take activities to resume and endpoint from an input file
glite-es-activty-restart tells the server, specified as endpoint, to restart the activity identifiers; can take activities to restart and endpoint from an input file
glite-es-activty-wipe tells the server, specified as endpoint, to wipe the activity identifiers; can take activities to wipe and endpoint from an input file
glite-es-activty-cancel tells the server, specified as endpoint, to cancel the activity identifiers; can take activities to cancel and endpoint from an input file
glite-es-notify-service sends to the server, specified as endpoint, a notification message (see below for the kind of messages available)
glite-es-delegate-proxy delegates a proxy into the remote endpoint
glite-es-delegation-info asks about a pre-created delegation on a certain endpoint
glite-es-delegation-renew renew a pre-created delegation on a certain endpoint
Creating simple activities on ES based CEs (no Input/Output sandboxes)The commandglite-es-activity-create creates an activity on an ES based CE; it needs the specification of and endpoint where the creation request must be sent, and an XML file containing the activity description (described in ADL language: https://twiki.cern.ch/twiki/bin/view/EMI/EmiExecutionService![]() $ cat ~/JDLs/simple_activity.ad <?xml version="1.0" encoding="UTF-8" standalone="yes"?> <CreateActivities> <ActivityDescription> <ActivityIdentification> <Name>Sleeping job</Name> <Description>A sleep for 120 seconds</Description> <Type>single</Type> </ActivityIdentification> <Application> <Executable> <Path>/bin/sleep</Path> <Argument>120</Argument> </Executable> </Application> <Resources> <QueueName>creamtest2</QueueName> </Resources> </ActivityDescription> </CreateActivities>To submit this ADL file simple_activity.adl just issue the command (also command's output is shown):
$ glite-es-activity-create -e cream-05.pd.infn.it ~/JDLs/simple_activity.adl ***************************************** ActivityID = CR_ES146448265 ActivityMgrURI = https://cream-05.pd.infn.it:8443/ce-cream-es/services/ActivityManagementService Status = PREPROCESSING Status Attrs = {} Timestamp = Tue Mar 20 11:05:23 2012 Description = ETNSC = STAGEIN Dir = {gsiftp://cream-05.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam002/14/CR_ES146448265/ISB} SESSION Dir = {} STAGEOUT Dir = {gsiftp://cream-05.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam002/14/CR_ES146448265/OSB}Please note some useful information returned by the server in addition to the activity's identifier ( CR_ES146448265 ):
* Status = PREPROCESSING , saying an obvious thing: just after activity creation, its status is still in a preliminary state waiting to be run somewhere
* Timestamp = Tue Mar 20 10:16:23 2012 , information about timestamp of actual activity creation on the server
* STAGEIN/STAGEOUT that we can ignore right now; we will discuss them later
Monitoring activities on ES based CEsObtaining the status of one or more activities is as simple as issuing this command:$ glite-es-activity-status -e cream-10.pd.infn.it CR_ES146448265 ***************************************** ActivityID = CR_ES853695732 Status = PROCESSING_RUNNING Attributes = {APP_RUNNING} Timestamp = Tue Mar 20 11:05:28 2012 Description =After a while the user's activity will be finished: $ glite-es-activity-status -e cream-10.pd.infn.it CR_ES146448265 ***************************************** ActivityID = CR_ES853695732 Status = TERMINAL Attributes = {} Timestamp = Tue Mar 20 11:07:28 2012 Description = reason=0The user can specify more than one activity identifier as arguments of status command: $ glite-es-activity-status -e cream-10.pd.infn.it CR_ES918217695 CR_ES499424509 ***************************************** ActivityID = CR_ES499424509 Status = TERMINAL Attributes = {} Timestamp = Wed Mar 21 09:45:57 2012 Description = reason=0 ***************************************** ActivityID = CR_ES918217695 Status = TERMINAL Attributes = {} Timestamp = Wed Mar 21 09:45:48 2012 Description = reason=0If the user asks for the status of an acitivity identifier not present on the endpoint, he/she will get an error message like this: $ glite-es-activity-status -e cream-05.pd.infn.it CR_ES920948151 FOO ***************************************** ActivityID = CR_ES920948151 Status = TERMINAL Attributes = {CLIENT_STAGEOUT_POSSIBLE} Timestamp = Thu Mar 22 13:29:11 2012 Description = reason=0 ***************************************** ActivityID = FOO Message = Activity not found! Timestamp = Thu Mar 22 14:34:56 2012 Description = N/A FailCode = N/A Creating activities on ES based CEs, that need to move Input/Output sandboxesAutomatically stage-in/stage-out performed by the CEThe previous example was about a simple activity that doesn't involve sandboxes to move around; if the user needs to send one or more files to the CE, or needs that the CE sends files after activity termination, then a more complex activity must be written and sent to the CE; before to create the activity a delegation must be created on the CE (or "a proxy must be delegated on the CE"), because moving the sandbox files could imply to contact remote authenticated services, like GridFTP servers. To delegate his/her own proxy on the CE, the user has to issue this command:$ glite-es-delegate-proxy cream-10.pd.infn.it DelegationID = 04669318871724504(this command glite-es-delegate-proxy will be explained more deeply later).
The returned delegation identifier 04669318871724504 (that is an handle to the real delegated proxy residing on the CE cream-10.pd.infn.it , whose lifetime is equal to the user's proxy lifetime) must be inserted in the proper ADL's section, like in this example:
$ cat ~/JDLs/activity_files.adl <?xml version="1.0" encoding="UTF-8" standalone="yes"?> <CreateActivities> <ActivityDescription> <ActivityIdentification> <Name>CustomJob</Name> <Description>A job that to run needs a file to be sent and writes its elaboration on an output file to be retrieved.</Description> <Type>single</Type> </ActivityIdentification> <Application> <Executable> <Path>myjob.sh</Path> <Argument></Argument> </Executable> <Environment> <Name>MY_ENV</Name> <Value>"my env"</Value> </Environment> <Error>JobError.txt</Error> <Output>JobOutput.txt</Output> </Application> <Resources> <QueueName>creamtest2</QueueName> </Resources> <DataStaging> <OutputFile> <Name>JobError.txt</Name> <Target> <URI>gsiftp://cream-23.pd.infn.it//tmp/JobError.txt</URI> <DelegationID>04669318871724504</DelegationID> </Target> </OutputFile> <OutputFile> <Name>JobOutput.txt</Name> <Target> <URI>gsiftp://cream-23.pd.infn.it//tmp/JobOutput.txt</URI> <DelegationID>04669318871724504</DelegationID> </Target> </OutputFile> <InputFile> <IsExecutable>true</IsExecutable> <Name>myjob.sh</Name> <Source> <URI>gsiftp://cream-23.pd.infn.it//tmp/myjob.sh</URI> <DelegationID>04669318871724504</DelegationID> </Source> </InputFile> </DataStaging> </ActivityDescription> </CreateActivities>When the proxy is delegated, and the ADL does contain the delegation identifier associated to the files to move, the activity can be created: $ glite-es-activity-create -e cream-10.pd.infn.it ~/JDLs/activity_files.adl ***************************************** ActivityID = CR_ES292405349 ActivityMgrURI = https://cream-10.pd.infn.it:8443/ce-cream-es/services/ActivityManagementService Status = PREPROCESSING Status Attrs = {} Timestamp = Tue Mar 20 10:59:15 2012 Description = ETNSC = STAGEIN Dir = {gsiftp://cream-10.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam004/29/CR_ES292405349/ISB} SESSION Dir = {} STAGEOUT Dir = {gsiftp://cream-10.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam004/29/CR_ES292405349/OSB}In the last example, the activity represents a job to execute, and it is not a system executable (like the /bin/sleep of the previous example); the executable is built up by the user, myjob.sh , and must be staged out from a storage server (cream-23.pd.infn.it in this example; see again the <Source> node in the ADL above) running a GridFTP daemon, into the activity's directory in the CE. We do not enter into the job's details: simply note that the user's executable produces a standard output, and a standard error that will be redirected to dedicated files (the JobOutput.txt and JobError.txt specified in the ADL above). The standard output will contain the print of a particular user-defined environment variable (MY_ENV, see the ADL above) and an echo message; the standard error will contain an error message triggered by operations that the user cannot perform on the destination worker node.
In the ADL it is written that the two files JobOutput.txt and JobError.txt must be sent to the storage server (the same GridFTP server seen before for the executable stage-out) cream-23.pd.infn.it ; of course it could be a different storage server. The user can retrieve these output files by mean of his/her preferred GridFTP client. For example:
$ globus-url-copy gsiftp://cream-23.pd.infn.it/tmp/JobOutput.txt JobOutput.txt $ globus-url-copy gsiftp://cream-23.pd.infn.it/tmp/JobError.txt JobError.txt $ ls -l Job* -rw------- 1 dorigoa dorigoa 120 Mar 21 10:31 JobError.txt -rw------- 1 dorigoa dorigoa 73 Mar 21 10:31 JobOutput.txtPlease note that the destination CE cream-10.pd.infn.it has sent the two files JobOutput.txt and JobError.txt to the storage server cream-23.pd.infn.it using the proxy that the user delegated before (identified by the string 04669318871724504 , that has been put in the ADL file) for authentication. The delegated proxy must be valid when the stage-in occurs. Then it is up to the user to delegate a proxy into the CE, that has a lifetime long enough for the entire time of the job's life and files stage-in.
Stage-in performed by the userConsider this ADL:<?xml version="1.0" encoding="UTF-8" standalone="yes"?> <CreateActivities> <ActivityDescription> <ActivityIdentification> <Name>CustomJob</Name> <Description>A job that to run needs a file to be sent and writes its elaboration on an output file to be retrieved.</Description> <Type>single</Type> </ActivityIdentification> <Application> <Executable> <Path>myjob.sh</Path> <Argument></Argument> </Executable> <Environment> <Name>MY_ENV</Name> <Value>"my env"</Value> </Environment> <Error>JobError.txt</Error> <Output>JobOutput.txt</Output> </Application> <Resources> <QueueName>creamtest2</QueueName> </Resources> <DataStaging> <ClientDataPush>true</ClientDataPush> <OutputFile> <Name>JobError.txt</Name> <Target> <URI>gsiftp://cream-23.pd.infn.it//tmp/JobError.txt</URI> <DelegationID>04352064184726456</DelegationID> </Target> </OutputFile> <OutputFile> <Name>JobOutput.txt</Name> <Target> <URI>gsiftp://cream-23.pd.infn.it//tmp/JobOutput.txt</URI> <DelegationID>04352064184726456</DelegationID> </Target> </OutputFile> </DataStaging> </ActivityDescription> </CreateActivities>Note the new XML node <ClientDataPush> and the lack of the <InputFile> node. This ADL tells the CE that the ISB file (in this case the user's executable myjob.sh ) will be sent by the user by hand and the CE has to wait until this operation is performed.
$ glite-es-activity-create -e cream-10.pd.infn.it ~/JDLs/activity_files_push.adl ***************************************** ActivityID = CR_ES613269972 ActivityMgrURI = https://cream-10.pd.infn.it:8443/ce-cream-es/services/ActivityManagementService Status = PREPROCESSING Status Attrs = {CLIENT_STAGEIN_POSSIBLE} Timestamp = Wed Mar 21 13:36:55 2012 Description = ETNSC = STAGEIN Dir = {gsiftp://cream-10.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam004/61/CR_ES613269972/ISB} SESSION Dir = {} STAGEOUT Dir = {gsiftp://cream-10.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam004/61/CR_ES613269972/OSB}Note the status attribute CLIENT_STAGEIN_POSSIBLE : the CE is saying that it is waiting for the user to send the required ISB by hand (by mean of the usual GridFTP client for example). Also note that the CE is communicating the complete address where to send the ISB (and the protocol supported, gsiftp):
$ globus-url-copy myjob.sh gsiftp://cream-10.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam004/61/CR_ES613269972/ISB/myjob.shWhen the stage-in is finished, the user has to notify the service that the file shipment is complete: $ glite-es-notify-service -e cream-10.pd.infn.it CR_ES613269972:CLIENT-DATAPUSH-DONEThe activity is started by the CE and the OSB are available as usual where specified in the ADL ( cream-23.pd.infn.it ).
In this example a new command has been introduced: glite-es-notify-service , that will be described more deeply later.
Both stage-in and stage-out performed by the userConsider this ADL:<?xml version="1.0" encoding="UTF-8" standalone="yes"?> <CreateActivities> <ActivityDescription> <ActivityIdentification> <Name>CustomJob</Name> <Description>A job that to run needs a file to be sent and writes its elaboration on an output file to be retrieved.</Description> <Type>single</Type> </ActivityIdentification> <Application> <Executable> <Path>myjob.sh</Path> <Argument></Argument> <FailIfExitCodeNotEqualTo>0</FailIfExitCodeNotEqualTo> </Executable> <Environment> <Name>MY_ENV</Name> <Value>"my env"</Value> </Environment> <Error>JobError.txt</Error> <Output>JobOutput.txt</Output> </Application> <Resources> <QueueName>creamtest2</QueueName> </Resources> <DataStaging> <ClientDataPush>true</ClientDataPush> <OutputFile> <Name>JobError.txt</Name> </OutputFile> <OutputFile> <Name>JobOutput.txt</Name> </OutputFile> </DataStaging> </ActivityDescription> </CreateActivities>This ADL doesn't contain specification of <Target> node, which means that the CE doesn't have to send any files to any remote server; it will be up to the user to retrieve the JobOutput.txt and JobError.txt files. As before, just after activity creation, the CE will send back the client the paths for stage-in and stage out. Let's skip the description of stage-in of executabe myjob.sh already described above. The activity creation with this ADL will return (as usual):
$ glite-es-activity-create -e cream-05.pd.infn.it ~/JDLs/activity_files_push_pull.adl ***************************************** ActivityID = CR_ES920948151 ActivityMgrURI = https://cream-05.pd.infn.it:8443/ce-cream-es/services/ActivityManagementService Status = PREPROCESSING Status Attrs = {CLIENT_STAGEIN_POSSIBLE} Timestamp = Thu Mar 22 13:28:32 2012 Description = ETNSC = STAGEIN Dir = {gsiftp://cream-05.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam002/92/CR_ES920948151/ISB} SESSION Dir = {} STAGEOUT Dir = {gsiftp://cream-05.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam002/92/CR_ES920948151/OSB}This time, in addition to the "STAGEIN Dir", the user has to consider (and remember) the value of "STAGEOUT Dir"; when the activity will be finished (use glite-es-activity-status to check that), he/she will have to retrieve the two files by mean of the usual GridFTP client:
$ glite-es-activity-status -e cream-05.pd.infn.it CR_ES920948151 ***************************************** ActivityID = CR_ES920948151 Status = TERMINAL Attributes = {CLIENT_STAGEOUT_POSSIBLE} Timestamp = Thu Mar 22 13:29:11 2012 Description = reason=0 $ globus-url-copy gsiftp://cream-05.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam002/92/CR_ES920948151/OSB/JobOutput.txt JobOutput.txt $ globus-url-copy gsiftp://cream-05.pd.infn.it/var/cream-es_sandbox/dteam/CN_Alvise_Dorigo_L_Padova_OU_Personal_Certificate_O_INFN_C_IT_dteam_Role_NULL_Capability_NULL_dteam002/92/CR_ES920948151/OSB/JobError.txt JobError.txtWhen output file retrieve is finished the user should notify the server with the command: $ glite-es-notify-service CR_ES920948151:CLIENT-DATAPULL-DONE -e cream-05.pd.infn.itNow the status's attributes CLIENT_STAGEOUT_POSSIBLE disappeared: dorigoa@lxgrid05 14:46:46 ~/emi/creamui_emi2>stage/usr/bin/glite-es-activity-status CR_ES920948151 -e cream-05.pd.infn.it ***************************************** ActivityID = CR_ES920948151 Status = TERMINAL Attributes = {} Timestamp = Thu Mar 22 13:29:11 2012 Description = reason=0 Using an output file for the command
The activity creation command has the ability to write the activity identifier(s) into an output file. If this file already exists, the user has to make sure that the output file has been already used for the same endpoint of the current activity creation. En example is better than any further explanation (the special ADL file |