Paul,
I think the NM is set up now within the 1.2 alpha testbed here.
$ ldapsearch -x -H ldap://gppce06:2135 -b 'Mds-Vo-Name=ralpha,o=Grid'
provides some network information. Could you take a look to check it
looks okay.
Steve
-----Original Message-----
From: Paul D Mealor [mailto:[log in to unmask]]
Sent: Friday, May 17, 2002 12:21 PM
To: [log in to unmask]
Subject: Re: A bit of self criticism
On Tue, 14 May 2002, Andrew McNab wrote:
> On Tue, 14 May 2002, Peter Clarke wrote:
>
> > Given that NetCost is about to go out could we use this to blitz
getting
> > the
> > network monitoring running on at least the UK core sites as well as UCL
> > By running i mean fully visible to Frank and publishing in the MDS
> >
> > It makes sense to ask Paul to prioritise this for a while, and be on
hand to
> > help with any installation questions.
>
> Starting from the RPM's in
> http://datagrid.in2p3.fr/distribution/datagrid/wp7/RPMS/
> what do we need to do?
There are four RPMs there, three of which are monitoring tools and one has
the information providers which publish the data into the MDS (there are 2
versions of that RPM).
Monitoring tools
================
edg-iperf-1.0.0-1, edg-pinger-1.0.0-2, edg-udpmon-1.0.0-1
The monitoring tools - edg-iperf, edg-pinger, edg-udpmon - should be
installed on any machine, although we recommend that they get installed on
a dedicated monitoring machine so that measurements aren't affected by
other jobs. These packages have the following dependancies:
- apache package
- perl package
- wget package
- tcsh package
- gnuplot package
- file /usr/bin/ppmtogif (libgr-progs for RedHat 6.2 or
netpbm-progs for RedHat 7.1)
and the installation procedure for the three is pretty much the same, and
involves installing the RPM then setting up Apache to display the tools'
web interface.
use rpm -i edg-udpmon
add following lines in /etc/httpd/conf/httpd.conf file :
Alias /udpmon "/opt/edg/udpmon/html/"
ScriptAlias /udpmon-bin/ "/opt/edg/udpmon/cgi-bin/"
use rpm -i edg-pinger
add following lines in /etc/httpd/conf/httpd.conf file :
Alias /pinger "/opt/edg/pinger/html/"
ScriptAlias /pinger-bin/ "/opt/edg/pinger/cgi-bin/"
use rpm -i edg-iperf
add following lines in /etc/httpd/conf/httpd.conf file :
Alias /iperf "/opt/edg/iperf/html/"
ScriptAlias /iperf-bin/ "/opt/edg/iperf/cgi-bin/"
Then restart HTTPD: /etc/rc.d/init.d/httpd restart
Information providers
=====================
edg-netmon-info-provider-1.0.3-1
The information providers can be installed on any GRIS or GIIS which has
edg-info-main-1.0.2-2 or greater installed. The added schema elements in
edg-info-main-1.0.2-2 will break any older version *Ftree* servers which
try to cache data from it; this isn't a problem with Globus MDS. So, if a
site has progressed to edg-info-main-1.0.2-2 (ie for TB1.2 *sigh*) then it
can be installed anywhere, and if not then it must be installed on a
standalone MDS. WP7 *will*, I believe, set up a NetMon GIIS just for this
eventuality.
Dependances:
perl
perl-libwww-perl
edg-info-main
perl-URI
perl-HTML-Parser
Installation:
Assuming edg-info-main is installed.
Install edg-netmon-info-provider-1.0.3-1
Do mkdir /opt/edg/info/mds/sbin/tmp (to fix a bug)
in /etc/edg/info-mds.conf change the following:
NETMON_PRESENT=yes
NETMON_HOSTS=<monhost>,<monhost>,<monhost>
NETMON_TOOLS=pinger,iperfer,udpmon
NETMON_SITE=0
NETMON_NES=man.ac.uk
<monhost> should be the name of the host on which the monitoring tools
were installed, and there needs to be on instance of <monhost> for each
tool listed in NETMON_TOOLS (and one tool for each host).
The NETMON_NES is the Network Element which this host measures... and
should be the "domain name" of the site, as in man.ac.uk for Manchester,
or ucl.ac.uk for UCL.
If Globus MDS is installed, then that needs to be configured (in addition
to the standard stuff like setting GLOBUS_USER)
in /etc/globus.conf:
GRID_INFO_GRIS=yes
GRID_INFO_EDG=yes
When a netmon GIIS is set up, we will probably ask that
GRID_INFO_GRIS_REG_HOST=ccwp7.in2p3.fr
GRID_INFO_GRIS_REG_VO_NAME=netmon
are added to make the GRIS register. This would only be necessary on a
standalone GRIS.
Start the EDG MDS (and configurator) with
/etc/rc.d/init.d/edginfo-mds start
If you don't want to run the Ftree MDS then edginfo-mds can be stopped
now. It needs to be run because it acts as configurator for Globus MDS.
Start the Globus MDS with
/etc/rc.d/init.d/globus-mds start
I believe that that would be it.
Paul
>
> Cheers,
>
> Andrew
>
>
--
Paul Mealor
--+-> [log in to unmask] <-+--
<-+-- www.hep.ucl.ac.uk/~pdm/ --+->
| +44 (0)207 679 3044 <-+--
--+-> x33044 |
|