Tags:
, view all tags

WMS Test Plan

Unit tests

N/A

Deployment tests

Generic repository

  • epel.repo
  • EGI-trustanchors
  • sl.repo
  • sl-security.repo

Installation test

First of all, install the yum-protectbase rpm: yum install yum-protectbase.noarch

Then proceed with the installation of the CA certificates by issuing:

yum install ca-policy-egi-core

Install the WMS metapackage:

yum install emi-wms

After the definition of the site-info.def file configure the WMS:

/opt/glite/yaim/bin/yaim -c -s site-info.def -n WMS

At the end of the installation the various init script should be checked with all parameters (start | stop | restart | status | version) (TBD)

Update test

Starting from a production WMS add the patch repository then issue:

yum update

If necessary reconfigure the WMS:

/opt/glite/yaim/bin/yaim -c -s site-info.def -n WMS

At the end of the update the various init script should be checked with all parameters (start | stop | restart | status | version) (TBD)

Functionality tests

Features/Scenarios to be tested

WMS can be deployed into two modes:

  1. Using an LB server installed in the same machine (BOTH mode)
  2. Using an external LB server (PROXY mode)
Both scenarios should be tested.

Test job cycle (from submission to output retrieve)

Submit a job to the WMS service and when finished retrieve the output; a the end the final status of the jobs should be Cleared.

Submission can be tested using different type of proxy:

  • Proxy from different VO (TBD)
  • Proxy with different ROLE (TBD)
  • Delegated proxy retrieved from a MyproxyServer (TBD)
  • RFC 3820 compliant proxy (TBD)
Test job submission with the following type of jobs:

Normal Job
  • Test the complete cycle submitting to the two types of CE: lcg and Cream Implemented.
More different jdls can added in the future. In particular these attributes should be tested:
  • DataRequirements (with differents DataCatalogType) (TBD)
  • OutputData (TBD)
  • InputSandboxBaseURI, OutputSandboxDestURI and OutputSandboxBaseDestURI (TBD)
  • AllowZippedISB and ZippedISB (TBD)
  • ExpiryTime (TBD)
  • ShortDeadlineJob (TBD)

Perusal job

Job perusal is the ability to view output from a job while it is running. Implemented.

DAG job

Directed Acyclic Graphs (a set of jobs where the input/output/execution of one of more jobs may depend on one or more other jobs). Implemented.

  • Also the nodes should be in state Cleared
More different jdls can added in the future.

Parametric Job

Multiple jobs with one parametrized description. Implemented.

Collection Job

Multiple jobs with a common description. There are two ways to submit collection:

  • you can create a single jdl with all the jdls of node Implemented.
  • you can submit all the jdls stored in a directory (bulk submission) Implemented.

Parallel Job

Jobs that can be running in one or more cpus in parallel. Implemented.

  • One of the charactheristics of this type of jobs is the possibility to pass these parameters directly to the CE:
    • WHOLENODES
    • SMPGRANULARITY
    • HOSTNUMBER
    • CPUNUMBER
Jdls combining these parameters should be used. TBD

Delegation

  • There are two types of delegation: the automatic ones or you can delegate before submission. Submit jdls using both methods Implemented.

  • Make a delegation with an expired proxy. Command should fails. Implemented.

  • Submit with an expired delegation. Command should fails. Implemented.

Shallow and deep re-submission

There two type of resubmission; the first is defined deep occurs when the user's job has stardted running on the WN and then the job itself or the WMS JobWrapper has failed. The second one is called shallow and occurs when the WMS JobWrapper has failed before starting the actual user's job. Implemented.

Job List-match Testing

Test various matching requests Implemented.

With data

Test matchmaking using data requests TBD

  • You need to register a file on an SE, then try a list-match using a jdl like this one (as InputData put the lfn(s) registered before):
###########################################
#      JDL with Data Requirements         #
###########################################

Executable = "calc-pi.sh";
Arguments = "1000";
StdOutput = "std.out";
StdError = "std.err";
Prologue = "prologue.sh";
InputSandbox = {"calc-pi.sh", "fileA", "fileB","prologue.sh"};
OutputSandbox = {"std.out", "std.err","out-PI.txt","out-e.txt"};
Requirements = true;

DataRequirements = {
[
DataCatalogType = "DLI";
DataCatalog = "http://lfcserver.cnaf.infn.it:8085";
InputData = {"lfn:/grid/infngrid/cesini/PI_1M.txt","lfn:/grid/infngrid/cesini/e-2M.txt"};
]
};
DataAccessProtocol = "gsiftp";

The listed CEs should be the ones "close" to the used SE

Gang-Matching
If we consider for example a job that requires a CE and a determined amount of free space on a close SE to run successfully, the matchmaking solution to this problem requires three participants in the match (i.e., job, CE and SE), which cannot be accommodated by conventional (bilateral) matchmaking. The gangmatching feature of the classads library provides a multilateral matchmaking formalism to address this deficiency.

Try some listmatch using different expressions of Requirements which use the anyMatch() function: TBD

WMS Job Cancel Testing

Test the cancellation of these type of jobs (final status should be Cancelled):

  • Submit and cancel a normal job Implemented.

  • Submit a dag job and then cancel it (the parent) Implemented.

  • Submit a dag job and then cancel some of its nodes Implemented.

  • Submit a collection job and then cancel it (the parent) Implemented.

  • Submit a collection job and then cancel some of its nodes Implemented.

  • Cancellation of a Done job should fails. Implemented.

Prologue and Epilogue jobs

In the jdl you can specify two attributes prologue and epilogue which are scripts that are execute respectively before and after the user's job. Implemented.

Proxy renewal

  • Submit a long job with myproxyserver set using a short proxy to both CE (lcg and CREAM). Job should finishes Done (Success) Implemented.
  • Submit a long job without setting myproxyserver using a short proxy to both CE (lcg and CREAM). Job should finishes Aborted with reason "proxy expired" Implemented.

WMS feedback

This mechanism avoid a job to remain stuck for long time in queue waiting to be assigned to a worker node for execution. There are three parameters in the jdl that can be used to manage this mechanism:

  1. EnableWMSFeedback
  2. ReplanGracePeriod
  3. MaxRetryCount
The test should submit a lot of long jobs with short ReplanGracePeriod using a small number of resources, at the end of the test some jobs should be replanned (i.e. reassigned to different CEs). This can be evinced from the logging info of the jobs. (TBD)

Limiter mechanism

The WMS has implemented a limiter mechanism to protect himself from overload. This mechanism is based on different parameters anc can be configured inside wms configuration file. All these parameters should be checked and tested. (TBD)
 
Usage:/usr/sbin/glite_wms_wmproxy_load_monitor [OPTIONS]...
      --load1       threshold for load average (1min)
      --load5       threshold for load average (5min)
      --load15    threshold for load average (15min)
      --memusage    threshold for memory usage (%)
      --swapusage    threshold for swap usage (%)
      --fdnum       threshold for used file descriptor
      --diskusage    threshold for disk usage (%)
      --flsize    threshold for input filelist size (KB)
      --flnum    threshold for number of unprocessed jobs (for filelist)
      --jdsize    threshold for input jobdir size (KB)
      --jdnum    threshold for number of unprocessed jobs (for jobdir)
       --ftpconn    threshold for number of FTP connections
       --oper       operation to monitor (can be listed with --list)
      --list       list operation supported
      --show       show all the current values

Purging

There are differents purging mechanisms on the WMS:
  • SandboxDir purging done by a cron script using command: /usr/sbin/glite-wms-purgeStorage.sh
  • Internal purging which means removal of files used by the various daemons as temporary information store. This purging should be done by the daemons themself.
  • Proxy purging done by a cron script using the command: /usr/bin/glite-wms-wmproxy-purge-proxycache
All these mechanism should be tested, i.e. check if all the unused files are removed at the end of the job. (TBD)

Configuration file

The file /etc/glite-wms/glite_wms.conf is used to configure all the daemons running on a WMS. A lot of parameters should be set with this file. Almost all these parameters should be checked. (TBD)

It should be verified that in the configuration file /etc/glite-wms/glite_wms.conf there are these hard-coded values:

For the common section:

DGUser = "\${GLITE_WMS_USER}"
HostProxyFile = "\${WMS_LOCATION_VAR}/glite/wms.proxy"
LBProxy = true

For the JobController section:

CondorSubmit = "${CONDORG_INSTALL_PATH}/bin/condor_submit"
CondorRemove = "${CONDORG_INSTALL_PATH}/bin/condor_rm"
CondorQuery = "${CONDORG_INSTALL_PATH}/bin/condor_q"
CondorRelease = "${CONDORG_INSTALL_PATH}/bin/condor_release"
CondorDagman = "${CONDORG_INSTALL_PATH}/bin/condor_dagman"
DagmanMaxPre = 10
SubmitFileDir = "${WMS_LOCATION_VAR}/jobcontrol/submit"
OutputFileDir = "${WMS_LOCATION_VAR}/jobcontrol/condorio"
InputType = "jobdir"
Input = "${WMS_LOCATION_VAR}/jobcontrol/jobdir/"
LockFile = "${WMS_LOCATION_VAR}/jobcontrol/lock"
LogFile = "\${WMS_LOCATION_LOG}/jobcontoller_events.log"
LogLevel = 5
MaximumTimeAllowedForCondorMatch = 1800
ContainerRefreshThreshold = 1000

For the NetworkServer section:

II_Port  = 2170
Gris_Port = 2170
II_Timeout = 100
Gris_Timeout = 20
II_DN = "mds-vo-name=local, o=grid"
Gris_DN = "mds-vo-name=local, o=grid"
BacklogSize = 64
ListeningPort = 7772
MasterThreads = 8
DispatcherThreads = 10
SandboxStagingPath = "${WMS_LOCATION_VAR}/SandboxDir"
LogFile = "${WMS_LOCATION_LOG}/networkserver_events.log"
LogLevel = 5
EnableQuotaManagement = false
MaxInputSandboxSize = 10000000
EnableDynamicQuotaAdjustment = false
QuotaAdjustmentAmount = 10000
QuotaInsensibleDiskPortion = 2.0
DLI_SI_CatalogTimeout = 60
ConnectionTimeout = 300

For the LogMonitor section:

JobsPerCondorLog = 1000
LockFile = "${WMS_LOCATION_VAR}/logmonitor/lock"
LogFile = "${WMS_LOCATION_LOG}/logmonitor_events.log"
LogLevel = 5
ExternalLogFile = "\${WMS_LOCATION_LOG}/logmonitor_external.log"
MainLoopDuration = 5
CondorLogDir = "${WMS_LOCATION_VAR}/logmonitor/CondorG.log"
CondorLogRecycleDir = "${WMS_LOCATION_VAR}/logmonitor/CondorG.log/recycle"
MonitorInternalDir = "${WMS_LOCATION_VAR}/logmonitor/internal"
IdRepositoryName = "irepository.dat"
AbortedJobsTimeout = 600
GlobusDownTimeout = 7200
RemoveJobFiles = true
ForceCancellationRetries = 2

or the Workloadmanager section:

PipeDepth = 200
WorkerThreads = 5
DispatcherType = "jobdir"
Input = "${WMS_LOCATION_VAR}/workload_manager/jobdir"
LogLevel = 5
LogFile  = "${WMS_LOCATION_LOG}/workload_manager_events.log"
MaxRetryCount = 10
CeMonitorServices = {}
CeMonitorAsynchPort = 0
IsmBlackList = {}
IsmUpdateRate = 600
IsmIiPurchasingRate = 480
JobWrapperTemplateDir = "${WMS_JOBWRAPPER_TEMPLATE}"
IsmThreads = false
IsmDump = "${WMS_LOCATION_VAR}/workload_manager/ismdump.fl"
SiServiceName = "org.glite.SEIndex"
DliServiceName = "data-location-interface"
MaxRetryCount = 10
DisablePurchasingFromGris = true
EnableBulkMM = true
CeForwardParameters = {"GlueHostMainMemoryVirtualSize","GlueHostMainMemoryRAMSize","GlueCEPolicyMaxCPUTime"}
MaxOutputSandboxSize = -1
EnableRecovery = true
QueueSize = 1000
ReplanGracePeriod = 3600
MaxReplansCount = 5
WmsRequirements  = ((ShortDeadlineJob =?= TRUE) ? RegExp(".*sdj$", other.GlueCEUniqueID) : !RegExp(".*sdj$", other.GlueCEUniqueID)) && (other.GlueCEPolicyMaxTotalJobs == 0 || other.GlueCEStateTotalJobs < other.GlueCEPolicyMaxTotalJobs) && (EnableWmsFeedback =?= TRUE ? RegExp("cream", other.GlueCEImplementationName, "i") : true)

For the WorkloadManagerProxy:

SandboxStagingPath = "${WMS_LOCATION_VAR}/SandboxDir"
LogFile = "${WMS_LOCATION_LOG}/wmproxy.log"
LogLevel = 5
MaxInputSandboxSize = 100000000
ListMatchRootPath = "/tmp"
GridFTPPort = 2811
LBLocalLogger = "localhost:9002"
MinPerusalTimeInterval = 1000
AsyncJobStart = true
EnableServiceDiscovery = false
LBServiceDiscoveryType = "org.glite.lb.server"
ServiceDiscoveryInfoValidity = 3600
WeightsCacheValidity = 86400
MaxServedRequests = 50
OperationLoadScripts = [
jobRegister = "${WMS_LOCATION_SBIN}/glite_wms_wmproxy_load_monitor --oper jobRegister --load1 22 --load5 20 --load15 18 --memusage 99 --diskusage 95 --fdnum 1000 --jdnum 1500 --ftpconn 300"
jobSubmit = "${WMS_LOCATION_SBIN}/glite_wms_wmproxy_load_monitor --oper jobSubmit --load1 22 --load5 20 --load15 18 --memusage 99 --diskusage 95 --fdnum 1000 --jdnum 1500  --ftpconn 300"
RuntimeMalloc = "/usr/lib64/libtcmalloc_minimal.so"
]

For the ICE section:

start_listener  =  false
start_lease_updater  =  false
logfile  =  "${WMS_LOCATION_LOG}/ice.log"
log_on_file = true
creamdelegation_url_prefix  =  "https://"
listener_enable_authz  =  true
poller_status_threshold_time  =  30*60
ice_topic  =  "CREAM_JOBS"
subscription_update_threshold_time  =  3600
lease_delta_time  =  0
notification_frequency  =  3*60
start_proxy_renewer  =  true
max_logfile_size  =  100*1024*1024
ice_host_cert  =  "${GLITE_HOST_CERT}"
Input  =  "${WMS_LOCATION_VAR}/ice/jobdir"
job_cancellation_threshold_time  =  300
poller_delay  =  2*60
persist_dir  =  "${WMS_LOCATION_VAR}/ice/persist_dir"
lease_update_frequency  =  20*60
log_on_console = false
cream_url_postfix  =  "/ce-cream/services/CREAM2"
subscription_duration  =  86400
bulk_query_size  =  100
purge_jobs  =  false
InputType  =  "jobdir"
listeneristener_enable_authn  =  true
ice_host_key  =  "${GLITE_HOST_KEY}"
start_poller  =  true
creamdelegation_url_postfix  =  "/ce-cream/services/gridsite-delegation"
cream_url_prefix  =  "https://"
max_ice_threads  =  10
cemon_url_prefix  =  "https://"
start_subscription_updater  =  true
proxy_renewal_frequency  =  600
ice_log_level  =  700
soap_timeout  =  60
max_logfile_rotations  =  10
cemon_url_postfix  =  "/ce-monitor/services/CEMonitor"
max_ice_mem = 2096000
ice_empty_threshold = 600

It should then be verified that:

  • The attribute II_Contact of NetworkServer section matches the value of the yaim variable BDII_HOST
  • The attribute WMExpiryPeriod of WorkloadManager section matches the value of yaim variable WMS_EXPIRY_PERIOD
  • The attribute MatchRetryPeriod of WorkloadManager section matches the value of yaim variable WMS_MATCH_RETRY_PERIOD
  • The attribute IsmIiLDAPCEFilterExt of WorkloadManager section is (|(GlueCEAccessControlBaseRule=VO:vo1)(GlueCEAccessControlBaseRule=VOMS:/vo1/*)(GlueCEAccessControlBaseRule=VO=vo2...
  • The attribute LBServer of the WorkloadManagerProxy section matches the value of yaim variable LB_HOST

Performance tests

Collection of multiple nodes

Submit a collection of n (a good compromise should be 1000) nodes. (TBD)

Stress test

Stress tests can parametrized some features: (partially implemented)

  • Type of submitted job (e.g. Collections, normal, dag, parametric)
  • Submission frequency (i.e. number of submissions for minute)
  • Number of submission (i.e. duration of test)
  • Number of parallel submission threads (i.e. each one with a different user proxy)
  • With or without automatic delegation
  • With or without resubmission enable
  • With or without proxy renewal enable
  • Jdl description
    • with or without sandbox
    • with or without cpu computation
    • executable duration
    • with or without data transfer
    • etc...
This could be an example of stress test
  • 2880 collections each of 20 jobs (2 days of test)
  • One collection every 60 seconds
  • Four users
  • Use LCG-CEs and CREAM-CEs (with different batch systems)
  • Use automatic-delegation
  • The job is a "sleep random(666)"
  • Resubmission is enabled
  • Enable proxy renewal

Regression tests

Complete list of Rfc tests

Note

Implemented. means that an automatic test exists. Otherwise test must be developed and or execute by hand.

Edit | Attach | PDF | History: r18 | r9 < r8 < r7 < r6 | Backlinks | Raw View | More topic actions...
Topic revision: r7 - 2011-10-20 - MassimoSgaravatto
 
  • Edit
  • Attach
This site is powered by the TWiki collaboration platformCopyright © 2008-2024 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback