On Fri, 15 Feb 2008 10:29:32 -0500 Lon Hohberger <lhh@xxxxxxxxxx> wrote: > On Fri, 2008-02-15 at 10:13 +0200, Pavlos Parissis wrote: > > > It works now and I know why it didn't work before. > > Actually, it's luck... see below. ;) > > > The last time I saved a copy of the ip.sh under /usr/share/cluster, > > reading you comment bellow I can guess that the backup file was used > > and not the patched file, even the backup file had a name ip.sh-bak. > > It's because readdir() doesn't necessarily read things in alphabetic > sort order. So, if the directory entry for 'ip.sh-original' came before > 'ip.sh', 'ip.sh-original' would get picked as the provider for the 'ip' > resource. > > However, if you remove a file or two and create new ones, there's a > chance that the directory entry might come after 'ip.sh'... making > things work! > > The rule to remember is that we won't use anything in /usr/share/cluster > that does not have the execute permissions bit set, so you can create > backups in /usr/share/cluster -- just 'chmod -x' them :) > > The theory for this mode of operation is that we don't have to impose > file name extension requirements on users and developers (for example, > SAP* have no extensions - they were developed by the open source > community)... I prefer to have restrictions and requirements set, make life more predictable... > > I am sorry but I did expect from the cluster to pick up the backup file. > > ...The downside is that it traps unsuspecting users - such as > yourself. :) > > Maybe we should put a 'README' file in /usr/share/cluster? (Or at least > a FAQ entry on the wiki?) Both will be better - README files are faster to be found and wiki pages are better readable Thanks for the explanation Pavlos -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
