Ivan Pantovic wrote: > > I have a straightforward question, > > is it possible to screw-up cluster node(s) by issuing ntpdate while > cluster is running? > > I had random fencing of nodes (usually when load is low) and ... > recently I discovered that nodes are syncronised to ntp server using > crontab and ntpdate. > I know that ntpdate is all but not "gentle" and should be used only on > system startup for "rude", prior to ntpd start, syncronisation. > > I could go trough heartbeat routines and try to figure it out but just > if someone already knows and is willing to share it? It shouldn't be. cman uses the kernel jiffies for timing information rather than actual wall-clock time. Having said that I haven't tested this, and some of the userland code might be more subject to clock variances. Mucking around with computer clocks too often is usually a bad idea ;) -- patrick -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
