SG GLITE-3 0 Assessment Notes

From EGEE-see WIki

Jump to: navigation, search

Contents

Introduction

A number of SEE-GRID sites will perform a short assessment of gLite-3.0 middleware. Here are the notes produced by these sites during the upgrade, as well as the results of their assessment efforts, including local testing, SFT results assessment, GStat results, stability and reliability of services during the daily operations.

BA-01-ETFBL Notes

When installing lcg-CE_torque and glite-UI on the same node, there are problems with RPM dependency conflicts related to fetch-crl issue (documented on ROLLOUT list).

Solution is to first do:

install_node <site-info.def> lcg-CE_torque

and then

install_node <site-info.def> glite-UI

Ignore removal of fetch-crl, and then proceed with configure_node (of course, for both node types at the same time, as usual).

Problems with old version of glite-config. For some reason, on three nodes apt-get dist-upgrade did not upgrade glite-config to version 1.6.x needed for configure_node. If you encounter this error:

Configuring config_glite
Traceback (most recent call last):
  File "/opt/glite/etc/config/scripts/glite-wn-config.py", line 34, in ?
    from gLiteInstallerLib import ConfigParams
ImportError: cannot import name ConfigParams
gLite configuration script has returned nonzero return code
Error configuring config_glite 

Be sure to upgrade glite-config.

AL-02-FIE Notes

Some package are not intstalled/upgraded ,I don't know what problem it may have with apt-get or somethong else. the tipycal packages that are not upgraded are :

The following packages have been kept back

  glite-config (1.4.13-1 => 1.6.29-0)
  glite-data-util-c (1.1.6-1 => 1.2.0-2)
  perl-SOAP-Lite (0.55-sl3 => 0.60a-1.1.el3.rf)
  unixODBC (2.2.8-2.3.0.2 => 2.2.11-1_EGEE)

AL-01-INIMA Notes

Phase 1: migration of CE/SE

- executed

       /opt/glite/yaim/scripts/install_node \
       /opt/glite/yaim/al_inima_config/site-info.def \
       lcg-CE_torque glite-SE_classic glite-MON 2>&1
       /opt/glite/yaim/scripts/install_node \
       /opt/glite/yaim/al_inima_config/site-info.def glite-UI 2>&1

If all functions together there problem with fetch-crl (see http://wiki.egee-see.org/index.php/SG_GLITE-3_0_Assessment_Notes)

- problems with password of mysql server

    changed the password for mysql server using mysqld_safe

- configure_node

       /opt/glite/yaim/scripts/configure_node \
       /opt/glite/yaim/al_inima_config/site-info.def \
       CE_torque SE_classic MON UI

Problem with missing /opt/glite/etc/config/scripts/glite-ui-config.py Follow-up LCG-ROLLOUT Digest - 9 May 2006 to 10 May 2006 archive

       solution was to set following variable in glite-ui.cfg.xml:
       <py-ui.DefaultVo value="dteam"/>. I used value="seegrid".

- There is yet a minor problem with site-info.def (?) because GSTAT complaints: GlueHostOperatingSystemRelease: SL **NOTE** version should numeric values delimited by '.'

End Phase 1

TR-01-ULAKBIM Notes

MK-01-UKIM_II Notes

HR-01-RBI Notes

Upgraded services: CE (lcg flavor), SE, MON, WN.

Only a few minor problems:

  • rpm dependency conflicts (known fetch-crl problem)
  • yaim problems due to incorect old site-info.def
  • updates for some yaim functions because local users are in ldap

RO-01-ICI Notes

The site has a standard configuration with a classical SE and MON on different machines, there are no collocated services on nodes, and no core services like RB, BDII or LFC. The upgrade of CE, SE, MON, UI and WNs to gLite(LCG flavour) was smooth, without any issues. Following the upgrades notes there were no problems encountered. The new version of site-info.def from glite-yaim was used and as WNs were installed together with torque. The following:

/opt/glite/yaim/scripts/install_node site-info.def glite-WN glite-torque-client-config 

was issued after lcg-WN_torque was removed.

As with every upgrade one should not forget to keep the configuration files for glite-yaim in a different directory than default, and also backup maui.cfg as it will be overridden during upgrade.

After upgrade an incorrect number CPUs free or running jobs was noticed but this was due some local problems which were fixed. The middleware seems to be stable.

RO-03-UPB Notes

Installed an UI node on 11oct06 a fresh GLITE-3.0.2 (yaim 3.0.22) and had the following problem:

vi /opt/glite/etc/config/scripts/glite-ui-config.py 407 os.system('touch %s/.test > /dev/null 2&>1' % dir) should be 2>&1

On the same issue. Installed on one node, UI and WN /opt/glite/yaim/scripts/configure_node site-info.def glite-WN glite-UI And got /opt/gridice/monitoring/etc/gridice-role.cfg: No such file or directory Resolved it by installing gridice-sensor and edg-fabricMonitoring. (Not installed automatically)

After configuring glite-WN just like in the 3.0.2 guide, if using torque, the pbs wn didn't get configured. So we got: [root@gw01 log]# pbsnodes -a wn001.rogrid.pub.ro

    state = state-unknown,down
    np = 2
    properties = lcgpro
    ntype = cluster

glite-WN had to be glite-WN_torque

Personal tools