6 Replies Latest reply on Jul 22, 2011 8:28 AM by zman

    sdmcache and long running jobs


      I have a package that I intend to deliver via policy. The package requires a reboot, but we don't want to force the machine to reboot or present a UI to the user. Thus, I've designed the package to do what it needs to do, wait for a reboot to happen, then proceed with the rest of the package. Relying on the fact that policies will retry upon failure.


      One fly in the ointment I thought of while driving home was the clean up of the sdmcache folder. If the task ultimately takes longer than two days, in theory the files I need to continue the package could be deleted.


      Will files that are part of an active task be deleted per the cache clean up? Typing this out, I realize that the agent would probably just redownload the files before trying to run the pacakage agains, so it would probably still work..

        • 1. Re: sdmcache and long running jobs
          zman Master

          There are numerous ways to handle this, with the easiest way being changing the default Discard Period for the cache to a hire value. The registry key is:


          Discard Period
          the time is in seconds default is 172800 which is 48 hrs or two days


          If you change this make sure you do it before the job is ran. You could also install the files outside of the cache and delete them afterwards, more work than changing the Discard Period.

          • 2. Re: sdmcache and long running jobs

            Understood, thanks zman.


            I am correct, though, in thinking when the job retries it will make sure it has all the files in the package? The logs seem to indicate such behavior.

            • 3. Re: sdmcache and long running jobs
              Catalysttgj Expert

              Unless you have really tiny hard drives in your workstations, it would be a really good idea to bump up the multicast delete times as Zman points out. This time is really really too short now. Talking with our TAM in the past about this, this time is really like 1990's specs. The time should be greater now considering the size of hard drives has grown quite a bit in the past 10-20 years!


              At any rate, If you do change your cache delete time, don't forget about the fact that there are actually two times to increase as well. regular and the subnet rep time. You'll see them both in the same reg path.


              We increased ours to the maximum, which is only 1 year. Nothing will live longer than that in the cache. At least not currently at 9.02.


              Good luck!

              • 4. Re: sdmcache and long running jobs
                zman Master

                It should check the local cache first and if not there redownload, however, based on the size of the package there may be delays or bandwidth issues with downloading twice.

                • 5. Re: sdmcache and long running jobs
                  LANDave SupportEmployee

                  I would recommend adding an Enhancement Request to the Enhancement Request Portal to have the default SDMCACHE interval changed.



                  • 6. Re: sdmcache and long running jobs
                    zman Master

                    I would suggest that the interval be a custom setting in the client config. BTW, love the use of the word Enhancement, was hoping for a trifecta.