Re: [BProc] node status is always down

[Erik A. H. <er...@he...>]

On Tue, Jul 23, 2002 at 03:28:41AM -0500, Sadanand Kota wrote:
> Hi,
> I have installed bproc on 2 of my linux systems taking the RPM from
> Clustermatic.
> I am able to run bpmaster and bpslave succesfully. But when I check the
> status of the machines using bpstat, It always gives status as down.
> If I try /etc/beowulf/node_up 0, the ouput is
> 
> 	node_up: Setting system clock.
> 	error moving to node 0: Invalid argument
> 	Fatal error performing:  /usr/lib/beoboot/bin/bdate 0
> 
> (The same with /etc/beowulf/node_up 1)
> 
> Any idea how to change node status to up?

"down" means that the slave is not connected to the master.  That
basically means that the slave either isn't connected with TCP or it
hasn't sent the right magic cookies.

There's no way to change "down" to anything else manually.  Once the
slave connects, the state will change to boot and the maste rdaemon
will run /etc/beowulf/node_up.  If that exits with status 0, the state
will change to up.  Otherwise the state will change to error.
Manually setting the node state to "down" (with bpctl) will cause the
slave to be disconnected.

I'd check to make sure bpslave is actually connecting to the master.
Try running bpslave with -d to make sure it's not just exiting with
some error.  Also, check the system logs.

- Erik