Hi all, >> Now that we have CREAM-CEs querying an Argus server and even if gLexec >> on worker nodes is tested but not yet used in production, I am worried >> the the Argus server is becoming a SPOF (Single Point of Failure). >> >> Is it feasable to install more than one Argus server and make it so that >> CREAM and gLexec would be unaffected by the loss of one Argus server ? >> > In the CREAM configuration file (/etc/glite-ce-cream/cream-config.xml) > you can specify more than one PEP daemon endpoint. > As far as I know the list of urls is used by the PEP client for finding > the first available PEP daemon. > So I think you can at least configure a set of PEP daemons in HA; I > presume it is possible to distribute even the PDP service, but for what > concerns the PAP honestly I don't know. You have to ensure the different instances will return the same mappings, so there will be at least _one_ SPOF, viz. the gridmapdir... At CERN we have multiple Argus hosts behind a load-balanced alias, all mounting the single gridmapdir from an NFS server.