SG RGMA Installation Guide

From EGEE-see WIki

Jump to: navigation, search

The R-GMA installation follows the general EGEE glite middleware installation procedure:

https://twiki.cern.ch/twiki/bin/view/LCG/GenericInstallGuide310


The only difference for SEE-GRID is that SEE-GRID only sites (sites that are NOT in EGEE production) *must* use SEE-GRID registry/schema server gserv1.ipp.acad.bg.

All sites must register their MON boxes in HGSM.


Once you have decided on the name of your MON node, on all of your nodes you must edit the following two lines in site-info.def:

MON_HOST=mon.$MY_DOMAIN
REG_HOST=lcgic01.gridpp.rl.ac.uk

by inserting the name of your MON node as MON_HOST and by setting (non-EGEE sites only!)

REG_HOST=gserv1.ipp.acad.bg

Then you must reconfigure the nodes. Note that installation of MON box on the same node as other services is not recommended and not supported.

The other two relevant lines are:

MYSQL_PASSWORD=set_this_to_a_good_password
APEL_DB_PASSWORD="APELDB_PWD"

where you must put two good passwords instead of the defaults.

In order to check if authentication works as expected, you can try

mysql -u accounting -h MONBOXHOSTNAME accounting

from your CE and also from your MON. If you have authentication problem, you must grant the appropriate privileges by logging in as root to the mysql on your MON box.

Please note that ports 8443 and 8088 must be open for incoming connections. Lines like this can do the trick:

-A INPUT -p tcp -m tcp --dport 8443 -j ACCEPT
-A INPUT -p tcp -m tcp --dport 8088 -j ACCEPT


Please do not run rgma-gin on your nodes, since the functionality provided by it is NOT supported and thus not available for SEE-GRID sites.

You can check if your R-GMA is working by running

rgma

command from a UI, after having generated a valid proxy. As a workaround you can run this command from other nodes, by copying a valid proxy and export X509_USER_PROXY to point to it.

After you have accomplished proper R-GMA installation, you can expect apel to publish accounting data. Check /var/log/apel.log on both CE and MON nodes, to see if you have any errors. Bear in mind that because of some peculiarities of apel it is safer to have the publishing hour somewhere after 4:00 UTC, and not before.

It has been found that sites using the pbs jobmanager have problems publishing. If the upgrade to the latest apel rpms does not solve their problem, they can try the following:


"The problem has been solved replacing old pbs.pm (on your CE) with new one (containing DGAS support) which can be found at http://glite.phy.bg.ac.yu/GLITE-3/AEGIS/pbs.pm . In order to force yaim to remember settings for future reconfigurations, you need to replace your /opt/globus/setup/globus/pbs.in with http://glite.phy.bg.ac.yu/GLITE-3/AEGIS/pbs.in ."


Additional information:

1. http://goc.grid.sinica.edu.tw/gocwiki/ApelFaq has extensive explanations

2. If you see: "Unknown attribute: /ApelConfiguration/DBProcessor@inspectTable" and an upgrade/reconfiguration does not solve the problem, look for the quick fix at: https://perseus.e-technik.uni-dortmund.de/index.php/GRID_MON_Fehlermeldungen 3. The publishing of accounting information happens from the CE to the MON, and then from MON to R-GMA. If you see in /var/log/apel.log on your CE:

/opt/glite/bin/apel-pbs-log-parser: line 29: /usr/java/j2sdk1.4.2_15/bin/java: No such file or directory then probably you have a configuration problem.


Good luck,

Emanouil Atanassov

Personal tools