Devel14 Work Log

2007-10-08 (Ale)

30% of jobs fail due to this problem:

Status Reason:      LBProxy is enabled
Unable to query LB and LBProxy
edg_wll_QueryEvents[Proxy]
Exit code: 110
LB[Proxy] Error: Connection timed out
(edg_wll_gss_connect())

The jobs stay in the waiting state forever and they can be "resubmitted" through the command: glite-wms-job-submit --start _jobid_

2007-10-05 (Ale)

Found a bug on wmproxy: logging the Enqueued/OK events the "queue" parameter is not set
- The symptom is: Status Reason: cannot retrieve jdl
- The real problem: glite-lb-proxy[8426]: Message incomplete (Message incomplete, missing or empty field DG.ENQUEUED.QUEUE.)

2007-10-04 (Ale)

Set jobdir ad input method for workload-manager instead of filelist
- Set these variables in the glite_wms.conf file (WorkloadManager section)
  - DispatcherType = "jobdir";
  - Input = "${GLITE_LOCATION_VAR}/workload_manager/jobdir";
- Create these directories and then give the ownership to glite.glite
  - /var/glite/workload_manager/jobdir/new/
  - /var/glite/workload_manager/jobdir/old/
  - /var/glite/workload_manager/jobdir/tmp/
- Restart the services

2007-10-03 (Ale)

Start to install a WMS 3.1

Use this repository: http://goldrake.cnaf.infn.it:8080/ibrido/etics-rpms_3_1_0_slc4/RPMS.wms/

Install yaim

rpm -ihv glite-yaim-core-4.0.0-14.noarch.rpm
rpm -ihv glite-yaim-wms-4.0.0-1.noarch.rpm

Configure yaim

Create site-info.def
Create users.conf
Create groups.conf

Resolve dependencies

bdii needs:
- yum install openldap-clients openldap-servers
- yum install perl-libwww-perl
  - perl-HTML-Parser
  - perl-HTML-Tagset
  - perl-URI

edg-mkgridmap needs:
- yum install perl-Crypt-SSLeay perl-DateManip perl-IO-Socket-SSL perl-LDAP perl-Net-SSLeay perl-Term-ReadKey perl-XML-DOM perl-XML-RegExp
  - perl-Convert-ASN1
  - perl-XML-NamespaceSupport
  - perl-XML-SAX

mysql needs:
- yum install perl-DBI perl-PlRPC perl-Net-Daemon

glite-wms-manager needs:
- rpm -ihv vdt_globus_info_essentials-VDT1.6.0x86_rhas_4-1.i386.rpm

condor needs:
- yum install compat-libstdc++-33

gridsite-apache (not present in the repository, why?) needs:
- yum install curl libidn

Install

/opt/glite/yaim/bin/yaim -i -s site-info.def -m glite-WMS

Post Install

Copy hostcert.pem hostkey.pem under /etc/grid-security/

Resolve dependencies (these are not required by rpm but are necessary)

Needs to configure edg-mkgridmap.pl (Patch #1403)
- yum install perl-XML-Parser

Needs to configure globus
- rpm -ihv gpt-VDT1.6.0x86_rhas_4-1.i386.rpm

Needs by lbproxy
- yum install mysql-server
  - perl-DBD-MySQL

Needs by wmproxy
- yum install httpd
  - apr
  - apr-util
  - httpd-suexec
- yum install mod_ssl
  - distcache
- rpm -ihv glite-security-lcmaps-plugins-voms-1.3.7-3.slc4.i386.rpm

needs by condor (gahp)
- rpm -ihv condor-lcg-1.1.0-1.i386.rpm

Needs by configuration script
- rpm -ihv glite-lb-logger (Why do you not install it?)

Bugs Fixes

Remove from file /opt/glite/yaim/node-info.d/glite-wms the line:
- config_gip_service (why do you configure a GIP on a WMS?)

Fix line 41 of /opt/glite/yaim/functions/config_glite_lbproxy
- /etc/init.d/mysqld start

To start correctly workload-manager add to file /opt/glite/yaim/functions/config_glite_wms (it needs liblog4cxx.so.9.0.0)
- yaimgridpath_append LD_LIBRARY_PATH /usr/local/lib

To run correctly wmproxy add to file /opt/glite/yaim/functions/config_glite_wms (for libfcgi* and libtar)
- yaimgridpath_append LD_LIBRARY_PATH /opt/fcgi/lib
- yaimgridpath_append LD_LIBRARY_PATH /opt/libtar/lib

Configure

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

Post Configure

Install voms certificate:
- rpm -ihv lcg-vomscerts-4.6.0-1.noarch.rpm

NOTE

/opt/glite/bin/glite-lb-interlogd doesn't exit with a "/opt/glite/etc/init.d/glite-lb-locallogger stop" you need "Kill -9"
you need to configure LBServer = {"devel15.cnaf.infn.it:9000"}; on glite_wms.conf
/opt/glite/bin/glite_wms_wmproxy_dirmanager is bugged due to libtar link (Bug under investigation by Maraska) use this one
chmod 4555 /opt/glite/bin/glite_wms_wmproxy_dirmanager (Bug under investigation by Maraska)
use the correct glite-wms-purger.cron script
set the right value HostProxyFile = "${GLITE_LOCATION_VAR}/wms.proxy"; on glite_wms.conf
install rpm -ihv lcg-expiregridmapdir-2.0.0-1.noarch.rpm and fix lcg-expiregridmapdir cron job (with the correct path)
install glite-wms-wmproxy-logrotate.cron and set its conf file (/opt/glite/etc/wmproxy_logrotate.conf)
add MaxOutputSandboxSize = -1; to glite_wms.conf (see bug #27215)

Installing the new gridftp server (vdt1.6)

Install these rpms:
- vdt_globus_data_server-VDT1.6.0x86_rhas_4-3.i386.rpm
- glite-security-lcas-lcmaps-gt4-interface-0.0.13-2.slc4.i386.rpm
- glite-security-lcas-1.3.7-1.slc4.i386.rpm
- glite-security-lcas-plugins-voms-1.3.4-3.slc4.i386.rpm
- glite-security-lcas-interface-1.3.6-2.slc4.i386.rpm
- glite-security-lcas-plugins-basic-1.3.2-2.slc4.i386.rpm

Remove edg-mkgridmap from the cron.d dir

Add the proper VO entry in the /etc/grid-security/grid-mapfile: e.g.

Add the proper VO entry in the /etc/grid-security/groupmapfile: e.g.

Configure the gridftp server:
- Run /opt/glite/sbin/gt4-interface-install.sh install (it creates /etc/grid-security/gsi-authz.conf)
- Touch the file /opt/glite/etc/lcas/ban_users.db
- Create the file /opt/glite/etc/lcas/lcas.db: e.g.
- Create the file /opt/glite/etc/lcmaps/lcmaps.db.gridftp: e.g.

To start the service:
- export LCMAPS_DB_FILE=/opt/glite/etc/lcmaps/lcmaps.db.gridftp
- /opt/globus/sbin/globus-gridftp-server -p 2811 -l /var/log/gridftp.log -S

Remove old gridftp service

rpm -e --nodeps glite-wms-thirdparty-globus_gridftp_server
rpm -e --nodeps edg-mkgridmap

rpm -e perl-Crypt-SSLeay perl-DateManip perl-IO-Socket-SSL perl-LDAP perl-Net-SSLeay perl-Term-ReadKey perl-XML-DOM perl-XML-RegExp
rpm -e perl-Convert-ASN1 perl-XML-NamespaceSupport perl-XML-SAX

-- AlessioGianelle - 03 Oct 2007

Topic revision: r7 - 2007-10-08 - AlessioGianelle

Account
- Log In

~~Edit~~
~~Attach~~