Deployment Solution

We use DS 6.9. Sometimes we get problems when deploying to multiple PC's. We had this problem yesterday. I deployed an image to 20 PC's and ended up with between 8 and 10 masters depending on when the deployment console updated the display. It looked like some of the PC's were running more than one download at the same time. In fact some of the PC's were alternating between master and client. Naturally with all these masters and clients running the times were in the hours. I left them running and eventually after a few hours they all finished successfully. Does anyone have a clue as to what's happening? Normally this job would take 45mins, usually works fine and selects only one master.

Please provide your multicast settings as it is this functionality that uses masters. Also let us know what imaging engine you are using (rdeploy, ghost or imagex)?

The advice above is good, but we'll get you a head-start on figuring it out until you get that update in.

There are a number of reasons to have multiple masters during a multicast session (not all of them good and/or ideal) including:

• multiple LAN segments.  There has to be a master on each segment, so if those 20 systems were spread out, for instance, all over the globe, then yes you'll get multiple masters.
• Failed multicasts for any reason.  IF a slave falls behind too far, the master will literally cut them off to allow the rest of the multicast session to continue.  More precisely, if the packets the slave is asking for are no longer in RAM on the master, the slave is let go because the master can't "back up" to re-request the imag.
• Too low of thresholds OR too high with too short of timeouts.  IF the threshold is 2, and 3 are connected, those 3 may start their own session.  OR, if 10 is the threshold and 8 connect prior to the timeout, they'll all unicast.

These are just some reasons.  Performance affects the 2nd bullet, and depending on the tools used, that can be very real.

IF you have the bandwidth BTW, on that low of count, many of our customers literally prefer to simply unicast to every one of them.

Thanks for your interest in this problem. Multicast threshold 1. We use Rdeploy with a range of 8 multicast addresses and 5 ports. All clients have WinPE 2.1 partitions. The PC's are all on the same VLAN/segment and are connected by Cisco Catalyst 4506 switches running at 100Meg. Unicasting more than 6 or 7 PC's at a time is not an option as this grinds to a virtual halt. Also, more than 6 or 7 masters grinds to a virtual halt. Multicasting to 60 or 70 PC's normally works fine and usually only uses 1 master.

Threshold of 1 is too low.  That tells the master what the minimum count needs to be before it should kick off a multicast.  Try upping that to 10 and let us know how things pan out, OK?

Thanks for your suggestion. I will change the threshold and test it out. However this may cause problems when deploying 6 or 7 PC’s as they would all become masters and consequently run very slow. If this does prove to be the problem I may need to investigate other values, say 5, to try and get the best of both worlds. I'll let you know what happens.

That seems to have fixed it. I'll now have to find the most suitable threshold. Thanks for your help.

No problem.  It is difficult to find the right threshold.  This is another reason why, IF there's bandwidth, some prefer simply unicasting.  But at least you're now on the right track.  Good luck!!!