CMSSW Generic File Monitoring

Starting with CMSSW_6_1_0, the Xrootd file-close monitoring has been implemented as a CMSSW framework service. This allows for applications like the CERN popularity service to monitor all file accesses done by CMSSW applications and provides the experiment with a more complete view of what files are used and how they are used.

This service is off by default and can be enabled on a per-job basis (see "edmPluginHelp -p SiteLocalConfigService" for details) or a per-site basis. Enabling the service is done by modifying the site-local-config.xml.

For each file read by CMSSW, a single UDP packet is sent to the monitoring host. We believe this system is scalable for the foreseeable future.

Enabling File Monitoring

To enable file monitoring, add the following tag inside the <source-config> tag of your site-local-config.xml:

      <statistics-destination name="cms-udpmon-collector.cern.ch:9331" />

If there is no <source-config> tag in your site-local-config.xml, add one inside the tag.

Example site-local-config.xml

Here is a complete site-local-config.xml, including the statistics destination line:

<site-local-config>
 <site name="T2_US_Nebraska">
    <event-data>
      <catalog url="trivialcatalog_file://cvmfs/cms.cern.ch/SITECONF/T2_US_Nebraska/PhEDEx/storage.xml?protocol=direct"/>
      <catalog url="trivialcatalog_file://cvmfs/cms.cern.ch/SITECONF/T2_US_Nebraska/PhEDEx/storage.xml?protocol=xrootd"/>
    </event-data>
    <source-config>
      <statistics-destination name="cms-udpmon-collector.cern.ch:9331" />
    </source-config>
    <local-stage-out>
      <command value="hadoop" />
      <catalog url="trivialcatalog_file://cvmfs/cms.cern.ch/SITECONF/T2_US_Nebraska/PhEDEx/storage.xml?protocol=hadoop"/>
      <se-name value="srm.unl.edu" />
    </local-stage-out>
    <fallback-stage-out>
      <se-name value="srm-dcache.rcac.purdue.edu"/>
      <lfn-prefix value="srm://srm-dcache.rcac.purdue.edu:8443/srm/managerv2?SFN=/"/>
      <command value="srmv2"/>
    </fallback-stage-out>
    <calib-data>
        <frontier-connect>
        <load balance="proxies"/>
        <proxy url="http://red-squid1.unl.edu:3128"/>
        <server url="http://cmsfrontier.cern.ch:8000/FrontierInt"/>
      </frontier-connect>
    </calib-data>
 </site>
</site-local-config>

Any support questions may be sent to the hn-cms-wanaccess Hypernews forum.

Sample Information Packet

Below is a sample information packet from a CMSSW job. Note the DN information is only recorded if present; this is the same information collected by Frontier's monitoring system.

{"site_name":"T2_US_Nebraska",
 "fallback": true,
 "user_dn":"/DC=com/DC=DigiCert-Grid/O=Open Science Grid/OU=People/CN=Brian Bockelman",
 "client_host":"brian-test",
 "client_domain":"unl.edu",
 "server_host":"cmsstor359",
 "server_domain":"fnal.gov",
 "unique_id":"8ABDCAFE-B469-E211-82E9-00163ED711AD-0",
 "file_lfn":"/store/relval/CMSSW_6_1_0-GR_R_61_V6_RelVal_wEl2012C/SingleElectron/RECO/v1/00000/FAC88284-414C-E211-84E7-002618943947.root",
 "file_size":2198266271,
 "read_single_sigma":630363,
 "read_single_average":161060,
 "read_vector_average":1.26992e+07,
 "read_vector_sigma":7.91128e+06,
 "read_vector_count_average":21.3929,
 "read_vector_count_sigma":70.4056,
 "read_bytes":358477665,
 "read_bytes_at_close":358477665,
 "read_single_operations":18,
 "read_single_bytes":2899080,
 "read_vector_operations":28,
 "read_vector_bytes":355578585,
 "start_time":1359423686,
 "end_time":1359423788
}

This packet is under 900 bytes in size.

Edit | Attach | Watch | Print version | History: r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r2 - 2014-05-13 - JohnArtieda
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback