Difference between revisions of "Tools/Manuals/TS191"
Latest revision as of 12:41, 23 November 2012
|Main||EGI.eu operations services||Support||Documentation||Tools||Activities||Performance||Technology||Catch-all Services||Resource Allocation||Security|
|Documentation menu:||Home •||Manuals •||Procedures •||Training •||Other •||Contact ►||For:||VO managers •||Administrators|
Back to Administration FAQ
How to publish different memory limits for different queues on the same CE
By default YAIM only supports only very simple batch system configurations. Sites with more complicated configurations may want to set up multiple CEs with different configurations, e.g. submitting to different batch system subclusters or with different memory limits. Some sites have created local versions of some YAIM functions to allow any of their CEs to submit to any of their queues possessing properties that YAIM could not handle by default.
Since early 2011 a new node type is available to allow such cases to be handled in a standard way: glite-CLUSTER. It also has other benefits that are described here:
The glite-CLUSTER node can be used in conjunction with the LCG CE and the CREAM CE in EMI/UMD (a manual recipe is available for use in gLite 3.2).
Consult the documentation for further details:
Beware: the WMS currently cannot handle multiple GlueSubCluster objects referring to the same GlueCluster object in the information system: since it does not forward memory requirements etc. to the CE yet, it requires that the GlueSubCluster is uniquely determined by specifying the GlueCE, i.e. the queue, which implies specifying the GlueCluster. Therefore any GlueCE should refer to a GlueCluster that has a single GlueSubCluster. However, different queues may refer to different GlueCluster objects and thus to different GlueSubCluster objects, each with its own properties like memory limits, hardware type, OS, etc.