Fokke Dijkstra wrote:
> Hello Maarten,
>
> Thanks for the warning! I will try to get this static info into the other BDII as well.
>
> BTW isn't it strange that a CE publishes information that is needed by the SE? Apparently you cannot have a standalone SE at all.
Indeed, it came from the wish that a Classic SE be NFS-mountable on WNs,
which at this moment is no longer supported. We could reconsider it when
there is an SRM SE that exposes a complete POSIX file system API, but then
we would at the same time change the information system and the data
management tools such that there is no cross-talk between SE and CE.
We already discussed those changes and they seem easy, so we could get rid
of the GlueCESEBindCEAccesspoint requirement fairly soon, but at this time
the matter is not seen as high priority. A bug could be opened about it...
> LHC Computer Grid - Rollout wrote:
>
>>On Fri, 22 Apr 2005, Fokke Dijkstra wrote:
>>
>>
>>>Dear all,
>>>
>>>Next week we will upgrade our cluster to LCG-2.4.0. At this
>>
>>point in time we will also upgrade from RedHat 7.3 to
>>Scientific Linux 3.0.4. This means that the upgrade will be a
>>lot of work, and we have planned the whole week (April 25-29) for it.
>>
>>>During the upgrade we hope to keep downtimes as short as
>>
>>possible. To prevent loss of jobs we need to shutdown the
>>queues on the CE however.
>>
>>>The upgrade will not involve the SE teras.sara.nl and the RLS
>>>mu5.matrix.sara.nl!
>>>
>>>Because the data of these services is published on the CE
>>
>>which will go down, we have installed a separate BDII on the
>>RLS server.
>>
>>>During the upgrade we may move the information about
>>
>>certain services to this BDII to prevent the loss of this
>>information.
>>
>>>So now SARA has the following BDIIs with site information:
>>>
>>># SARA-LCG2
>>>ldap://giis.matrix.sara.nl:2170/mds-vo-name=saralcg2,o=grid #
>>>SARA-RLS ldap://rls.matrix.sara.nl:2170/mds-vo-name=sararls,o=grid
>>>
>>>Note, that in order to have this information available to
>>
>>you, the new BDII should be included in the top BDII you use.
>>The "all sites" list currently has this BDII, so most sites
>>should be OK.
>>
>>>I'm not sure about experiment specific BDIIs though. So it
>>
>>may be wise to check this if you need to access data stored
>>at the teras SE.
>>
>>You need to ensure that the GlueCESEBindSEUniqueID DNs remain
>>published during the upgrade, otherwise the lcg-* tools
>>cannot use your SE. For example:
>>
>>--------------------------------------------------------------
>>----------------
>># teras.sara.nl,
>>mu6.matrix.sara.nl:2119/jobmanager-pbs-atlas, saralcg2, grid
>>dn:
>>GlueCESEBindSEUniqueID=teras.sara.nl,GlueCESEBindGroupCEUnique
>> ID=mu6.matri
>>x.sara.nl:2119/jobmanager-pbs-atlas,mds-vo-name=saralcg2,o=grid
>>objectClass: GlueGeneralTop objectClass: GlueCESEBind
>>objectClass: GlueSchemaVersion
>>GlueCESEBindCEAccesspoint: /home
>>GlueCESEBindCEUniqueID: mu6.matrix.sara.nl:2119/jobmanager-pbs-atlas
>>GlueCESEBindSEUniqueID: teras.sara.nl
>>GlueSchemaVersionMajor: 1
>>GlueSchemaVersionMinor: 1
>>--------------------------------------------------------------
>>----------------
>>
>>These entries normally come from the CE, but since they are
>>static, any BDII can publish them.
>
>
>
>
> --------
> Fokke Dijkstra
> High Performance Computing
> SARA - Reken- en Netwerkdiensten http://www.sara.nl
> Tel. +31 20 592 8004 Fax. +31 20 668 3167
|