Hi,
Yes, this is all that is needed for a completely separate cimpute
cluster.
My point abouut the accounting still holds. It probably worth all sites
that have more than one CE connecting to their accounting databases
occasionally and running a query like:
select count(RecordIdentity), SUBSTRING(ExecutingCE, 1, 9), EventDate
from LcgRecords group by Eventdate, SUBSTRING(ExecutingCE, 1, 9);
You should see an entry for every CE of every date (unless no jobs were
submitted through it that day for any reason).
The end current output for RALPP looks like:
| 364 | heplnx206 | 2008-10-24 |
| 3705 | heplnx207 | 2008-10-24 |
| 3 | heplnx206 | 2008-10-25 |
| 1763 | heplnx207 | 2008-10-25 |
| 1503 | heplnx207 | 2008-10-26 |
| 65 | heplnx207 | 2008-10-27 |
+-----------------------+------------------------------+------------+
1638 rows in set (59.59 sec)
where you can see that heplnx206 hasn't entered any records since the
25th and not many that day. Now we installed the latest gLite on that
machine on the 23rd and I obviously forgot to go back and correct the
fact my accounting database is not on my MON box, but jobs submitted
before then were being accounted when they finished by the PBS server
and so entries still show up for a few days after the CE stops
publishing.
Yours,
Chris.
> -----Original Message-----
> From: Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Alessandra Forti
> Sent: 27 October 2008 17:05
> To: [log in to unmask]
> Subject: Re: Installing a second CE - tips and tricks
>
> Hi Matt,
>
> you need to
>
> *) add the CE to the site BDII regions
> *) explicitely set the number of CPUs in both the services BDII.
> (GlueSubClusterPhysicalCPUs: in
> /opt/glite/etc/gip/ldif/static-file-Cluster.ldif)
> *) list the new CE in the GOCDB
>
> cheers
> alessandra
>
>
> --
> Alessandra Forti - NorthGrid Technical Coordinator
> http://www.hep.manchester.ac.uk/computing/tier2
>
> Well you'll still need a tray
>
--
Scanned by iCritical.
|