WN Working Group Test Rig Log

Steps taken to investigate the ideas of the WNWorkingGroup.

14th Dec 2008

Now review vtb-generic-62 as a Cluster Node.

• Need to get software tags working.
• This needs to work. lcg-tags --add --ce vtb-generic-62.cern.ch --vo dteam --tags VO-dteam-abc
• We need the following packages.
  • vdt_globus_data_server, vdt_globus_essentials
  • edg-mkgridmap, edg-mkgridmap-conf
  • lcg-expiregridmapdir
  • glite-initscript-globus-gridftp
  • glite-security-lcas-lcmaps-gt4-interface, glite-security-lcas
  • The YAIM function config_gip_software_plugin should move from glite-yaim-lcg-ce to glite-yaim-cluster
• Add the following YAIM functions to the glite-CLUSTER
  • config_sysconfig_globus
  • config_sysconfig_edg
  • config_crl
  • config_host_certs
  • config_users
  • config_mkgridmap
  • config_vomsmap
  • config_globus_gridftp
  • config_lcas_lcmaps_gt4
  • config_gip_software_plugin
  • config_gip_vo_tag
  • config_vomsdir
  • config_add_pool_env
• Despite trying hard here I still can't get gridftp to authorise me. Will need some help. There is nothing fundamentally wrong or new, I just cannot get it right.
• Moving onto publishing of the tags.
  • Classic style tags are located in /opt/edg/var/info/<VO>/. I suggest we put the new per SubCluster tags in /opt/glite/var/info/<SubCluster>/<VO>/.
  • Consequently the yaim function config_subcluster should be extended to create these directories. Do it by hand now.
    • mkdir -p /opt/glite/var/info/SteveSubCluster/dteam
    • mkdir -p /opt/glite/var/info/SteveSubCluster/ops
  • Install the old lcg-info-dynamic-software and check it works.
    • It works just fine but needs to be updated now BUG:45310 to support the new per SubCluster tags. This work is not pending on anything and could be deployed tomorrow.
• Try and get Steve Burke's Cluster RTE publish thing working.
  • Install RPM. glite-info-provider-service and glite-info-provider-service
    • In fact we have to take the newer that production glite-info-provider-service-1.0.3-0.noarch.rpm which is in certification now to get the RTEpublish stuff.
    • To the glite-CLUSTER YAIM target we must add the new attached config_info_service_rtepublish.
    • Now looking at this there are a couple of changes that might make this better. Submitted to Steve Burke. BUG:45313.

18th Nov 2008

Now set up vtb-generic-62 as a Cluster Node.

• Add a site BDII as well to the cluster node.
• I don't want to install a whole CE will install only what I need. Of course this will be a node type meta package eventually.
  • glite-yaim-core
  • glite-yaim-cluster
  • glite-info-provider-generic
  • glite-info-provider-ldap (?)
  • glite-version
  • bdii
  • glue-schema
  • glite-info-templates
• Not obvious what CLUSTER_STEVECLUSTER_CE_HOSTS is needed for but we will see.
• Why now SITE_Name and not SITE_NAME?
• A resource BDII needs to be running on a Cluster Node. i.e all the yaim functions for that need adding. i.e config_edgusers, config_bdii_only and config_gip_only.

17th Nov 2008

Carry on configuration of new cluster.

• Check what the GlueCEHostingCluster should be set to in the GlueCE.
• Role=NULL being stripped out even when set. Needed by WMS.
• A little confused by CE_InfoApplicationDir I don't see why this is needed or is sensible on a GlueCE. Apparently some OSG thing.

12th Nov 2008

Carry on configuration of new cluster.

• config_torque_submitter_ssh should run after config_torque_server since it calls pbsnodes and requires a running pbs_server.

11th Nov 2008

Reinstall on new hardware (vtb-generic-xx) using YAIMcluster_1.

• CE_SMPSIZE needs some thought, in YAIM this is effectivly JOB_SLOTS, needs to fixed up with what ever we decide to do in the end. Also this really needs to be a per node value.
• UniqueID Currently set by the user, need to think of a way of making them really unique (tbc).
• White space error in one random function. config_apel_pbs

      -       requires $1 INSTALL_ROOT BATCH_SERVER MON_HOST SITE_NAME CE_HOST \ $
     +       requires $1 INSTALL_ROOT BATCH_SERVER MON_HOST SITE_NAME CE_HOST \$
     

• On the CE /opt/glite/yaim/bin/yaim -v -s /root/site-info.def -n lcg-CE -n TORQUE_utils -n TORQUE_server returned okay but with the "-c" flag it gave an error. ERROR: CE_InfoApplicationDir is not set
• GlueCEInfoApplicationDir on a GlueCE makes little sense if I understand it and it does not appear to be in Glue 1.3 anyway. Checking.

26th May 2008

Reinstall lxb1914 and lxb2091, Update jpackage repositories, install java.

lxb2091

• yum install glite-WN glite-TORQUE_client

lxb1914

• yum install lcg-CE  glite-WN glite-TORQUE_client glite-Torque-utils glite-TORQUE_server
• rpm --oldpackage --nodeps -Uvh glite-yaim-lcg-ce-1.0.0-1.noarch.rpm glite-yaim-core-5.0.0-1.noarch.rpm
• /opt/glite/yaim/bin/yaim -v -s /root/site-info/site-info.def -n TORQUE_client -n TORQUE_server -n WN -n lcg-CE

Notes on Installation

1. CE_SMPSIZE ... Need to think about putting nodes in subclusters in yaim via nodes.conf file.

14th Feb 2008

• Add two new queues dteamc and opsc linked to two a new cluster and subcluster both called lxb2091.cern.ch-c.
• Configure queues and existing queues to submit to sl4 or slc4 hosts.
• qmgr -c 's n lxb1914.crn.ch properties += slc4'
• qmgr -c 's q dteamc resources_default.neednodes=slc4'
• Edit the two subclusters so that one is SL4 and one SLC4.. It is a lie but does not matter in the PPS.

14th Feb 2008

• Added the two resources to the CERN_PPS BDII.
• ldap://lxb1914.cern.ch:2170/mds-vo-name=resource,o=grid
• ldap://lxb2091.cern.ch:2170/mds-vo-name=resource,o=grid

13th Feb 2008

• Change the GlueCluster and GlueSubCluster name on lxb2091 to be something more sensible, i.e change it to lxb2091 over lxb1914.
  • Edit the following files to do this, also add the CERN_PPS site name while we are there.
  • /opt/glite/etc/gip/ldif/static-file-Cluster.ldif
• Remove the GlueCluster and GlueSubCluster from lxb1914.
  • rm -f /opt/glite/etc/gip/ldif/static-file-Cluster.ldif
• Change the GlueCluster and GlueSubCluster ids in the GlueCE objects on lxb1914 in:
  • /opt/glite/etc/gip/ldif/static-file-CE.ldif

21st Dec 2007

• Install the required packages on lxb2091 to publish GlueCluster, GlueSubCluster.
• yum install PPS-glite-TORQUE_utils
• yum install lcg-info-dynamic-software
• yum install glite-yaim-lcg-ce
• Run selected functions from an =lcg-CE.
  • # /opt/glite/yaim/bin/yaim -r -s /root/siteinfo/siteinfo.def -n lcg-CE -f config_edgusers
  • # /opt/glite/yaim/bin/yaim -r -s /root/siteinfo/siteinfo.def -n lcg-CE -f config_gip_only
  • # /opt/glite/yaim/bin/yaim -r -s /root/siteinfo/siteinfo.def -n lcg-CE -f config_gip_software_plugin
  • # /opt/glite/yaim/bin/yaim -r -s /root/siteinfo/siteinfo.def -n lcg-CE -f config_gip_vo_tag
  • # /opt/glite/yaim/bin/yaim -r -s /root/siteinfo/siteinfo.def -n lcg-CE -f config_bdii_only
  • Don't run config_gip_ce. This does to much, this will need to be broken up for a real deployment.
• cp the following files from lxb1914 to lxb2091 to configure them (normally created by config_gip_ce).
  • /opt/glite/etc/gip/ldif/static-file-Cluster.ldif

20th Dec 2007

• Check the status, lxb1914 is correctly publishing GlueCE, GlueCluster and GlueSubCluster objects.
  • ldapsearch -x -H ldap://lxb1914:2170 -b 'Mds-vo-name=resource,o=Grid'
• globus-job-run lxb1914.cern.ch/jobmanager-lcgpbs -q dteam is working.

14th Nov 2007 lxb1914

• #  yum install PPS-glite-TORQUE_client PPS-glite-WN PPS-glite-TORQUE_server PPS-glite-TORQUE_utils PPS-lcg-CE
• # /opt/glite/yaim/bin/yaim -c -s /root/siteinfo/siteinfo.def -n WN -n TORQUE_client -n TORQUE_server -n TORQUE_utils -n lcg-CE

26th Oct 2007 lxb2091

• # /opt/glite/yaim/bin/yaim -c -s /root/siteinfo/siteinfo.def -n WN -n TORQUE_client
• Configure YAIM's config_java to do nothing.
• # yum install PPS-glite-BDII PPS-glite-WN PPS-glite-TORQUE_client.
• This node will be batchworker and publish the Glue(Sub)Clusters.

22nd Oct 2007 lxb1914

• # yum install PPS-glite-BDII PPS-glite-WN PPS-glite-TORQUE_client
• # /opt/glite/yaim/bin/yaim -c -s /root/siteinfo/siteinfo.def -n WN -n TORQUE_client.

18th Oct 2007 lxb1914

• Installed with SL4. Installed java, PPS-glite-BDII-3.1.0-1, torque and maui.
• This will be a resource BDII for GlueCE as well as the pbs_server. For luck it is also a pbs_mom as well.
• Really hoping the SL4 gatekeeper appears soon in the PPS.

-- SteveTraylen - 18 Oct 2007