4 Replies Latest reply on Jan 2, 2014 5:26 AM by human

    Patching more than two hosts in parallel

    human Rookie

      Hi Folks,

       

      I'm new to Shavlik and am experimenting with patching a subset of management servers using Shavlik under the 30 day trial.

       

      When running a deployment on say 10 servers discovered via vsphere, which are all in the one machine group using the same deployment template, the servers will only patch 2 at a time. So patches won't start installing on subsequent machines until the entire cycle including reboots has finished on the 1st two servers. I'd like deployment, installation and reboot of machines to be in parallel across the whole machine group. As some of our production environments contain hundreds of servers of the same type that should be patched together, doing 2 at a time will exceed matintence windows.

       

      Is this a limitation of the trial or am I doing something wrong?

       

      Cheers.

       

      Edit: Sorry should add I want to deploy without agents if possible.

        • 1. Re: Patching more than two hosts in parallel
          human Rookie

          bump, no thoughts?

          • 2. Re: Patching more than two hosts in parallel
            Rookie

            I cannot speak for the 30 day trial, but my experience patching VMs is that several tasks run in parallel.

            1 of 1 people found this helpful
            • 3. Re: Patching more than two hosts in parallel
              rworlton Employee

              The confusion here probably stems from the process used to deploy patches.

               

              When deploying patches using a push from the console and setting the scheduled start time to now, protect uses the following process

              1. it connects to the first machine in the list

              2. copies files to the destination machine

              3. creates a schedule to deploy the targeted patches

               

              From there on the console server relenquishes control to each individual client  machine to manage the rest of the deployment, and to report status to the console.

               

              After the console completes scheduling a patch deployment it moves on to the next machine and repeats the process.  The fact that your seeing 2 machines patch is probabably due to the speed at which the filecopy occurs.  The larger the file copy the fewer machines you'll see patch at the same time.

               

              if you want to have more machines patching simultaniously, you have two options.

               

              1. Schedule the deployment to occur at some time in the future.
                1. When choosing this option, the actual deployment of the requird patches will occur as soon as the job is scheduled. No patches are installed, only the pre-work of copying the files over and setting the schedule is done on each target machine.
                2. let's say that you chose 30 min into the future, for 30 minutes the console would be pre-staging the patch job on each machine in the patch group. At the end of 30 minutes all machines that had been successfully contatcted by the core would start to patch. 
              2. Use a Distribution server
                1. When using a distribution sever, the file copy is also delegated to each client machine as they will be told to go get the file from the distribution server. This means that while the process to kick off is still a serial process, youve taken out the latency caused by the file transfer and more machines will start patching within a very short window.

              Hope that answers your question.

               

              Thanks,

              • 4. Re: Patching more than two hosts in parallel
                human Rookie

                Ok thanks for the reply. Just getting back to looking at this, we've now purchased a license and want to begin deployments ASAP. I think the 1st option is a better bet, bandwidth isn't an issue to the end-points so I don't think we can justify distribution servers. I'll have a play around with the schedule.