>> I'd rather have it simply triggered when we have any result other than >> either a sucessful copy *or* the copy is not attempted. That would >> also catch things like "server up, file found but unreadable" (happens >> when you are reading from an NFS server with root squashed). > > In this case, I don't think it is right that cfengine tries to do an end > run around NFS. these seem to be orthogonal issues -- either you copy > from the server or you use NFS, but not both??
I do indeed use both. We have a core NFS server that provides various resources; if Cfengine uses those resources, I generally copy them into /var/cfengine first in order to provide some ability to function if the NFS server or network becomes unavailable. I typically do cfengine copies from the server rather than have clients hit the NFS server directly so as to avoid having several hundred clients exercising the automounter in rapid succession, and also to allow hosts that are not NFS clients to be able to use the resources. I think the disagreement stems from an assumption that failover classes are used to try copies by other means (and that is indeed a fine use for them). However, I primarily use the failover classes to trigger a complaint mechanism that alerts a monitoring system. (IMHO, Cfengine [or any app] should not try to make assumptions about what the admin will do with information provided.) Once I know cfengine did not know what I expected it to do, I can go poke around and find the exact problem. _______________________________________________ Help-cfengine mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-cfengine
