This shows you the differences between two versions of the page.
Both sides previous revisionPrevious revision | Next revisionBoth sides next revision | ||
wiki:old:customization_tips [2014/08/29 16:48] – [oarsh completion] neyron | wiki:old:customization_tips [2015/04/03 15:14] – neyron | ||
---|---|---|---|
Line 1: | Line 1: | ||
====== Configuration ====== | ====== Configuration ====== | ||
//In this section, you'll find advanced configuration tips// | //In this section, you'll find advanced configuration tips// | ||
- | ===== Using oaradmin to initiate the resources ===== | ||
- | You can install oardamin by installing the **oar-admin** package or typing **make tools-install && make tools-setup** from the sources. | ||
- | |||
- | Example for a cluster with 10 nodes with 2 hexa-core processors per node; the name of the nodes are james1, james2,... james10: | ||
- | < | ||
- | | ||
- | </ | ||
- | |||
- | Example for an hybrid cluster (72 itaniums cores SMP and 28 xeon cores) | ||
- | < | ||
- | | ||
- | | ||
- | </ | ||
- | |||
- | Oaradmin only prints a set of " | ||
- | < | ||
- | | ||
- | </ | ||
- | |||
- | **Note:** oaradmin checks in the oar database if you have the necessary properties (for example " | ||
- | < | ||
- | oarproperty -a cpu | ||
- | oarproperty -a core | ||
- | </ | ||
- | |||
===== Priority to the nodes with the lower workload ===== | ===== Priority to the nodes with the lower workload ===== | ||
This tip is useful for clusters of big nodes, like NUMA hosts with numerous cpus and a few nodes. When the cluster has a lot of free resources, users often wonder why their jobs are always sent to the first node while the others are completely free. With this simple trick, new jobs are sent preferably on the nodes that have the lowest 15 minutes workload. | This tip is useful for clusters of big nodes, like NUMA hosts with numerous cpus and a few nodes. When the cluster has a lot of free resources, users often wonder why their jobs are always sent to the first node while the others are completely free. With this simple trick, new jobs are sent preferably on the nodes that have the lowest 15 minutes workload. |