1 2 3 4 Previous Next 32 Replies Latest reply on Feb 15, 2010 8:06 PM by mrepine

    AutoD - DumpMachineUsers Question

      For the past couple months I've been noticing in AutoD log a long delay for the DumpMachineUsers command to complete.  See below.

       

      2/15/2010 10:59:37.78 AM:  Getting the current list of users for machine "Machinexx"
      2/15/2010 10:59:37.78 AM:  Running sbadmclDumpMachineUsers (Machinexx)...
      2/15/2010 10:59:37.88 AM:  Existing Connection
      2/15/2010 11:05:38.58 AM:  0x00000000 | The operation completed successfully.
      2/15/2010 11:05:38.58 AM:  0x00000000 | The operation completed successfully.

       

      So my question is what would cause this to take so long to execute?  My understanding is it's dumping the users assigned for the machine and since there are only 5 users assigned it shouldn't take that long.  When we first started deploying machines there was no delay.  Now that we have 2000 user and 2000 machines it's taking on average about 5 minutes to complete.  Indexing is enabled. Is this a normal amount of time for the command to execute based on the current number of machines and users?  This is also causing the same delay after the first reboot for setting up personalized settings.  The CPU utilization on the DB server was less than 10% when I ran this test.  Plenty of memory available.  The client and the DB server are on the same network segment.  Currently running v5.1.7.  Any help with would be appreciated.

        • 1. Re: AutoD - DumpMachineUsers Question
          safebootsamurai

          It's my understanding that the code is set to search all machine groups for your machine then dump the user list for that machine, if it were setup to search only that machines group it would go faster.

           

          You can test by running the command both ways on the client.

           

          <Autodomains Logic for DumpMachineUsers>

          Set command = xmldom.selectSingleNode("SafeBoot/SbAdminScripting/SbAdminCommand")

          command.selectSingleNode("Command").text = "DumpMachineUsers"

          command.selectSingleNode("File").text = tname

          command.selectSingleNode("ExpandGroups").text="True"

          command.selectSingleNode("Machine").text= myMachineName

          command.selectSingleNode("Group").text="*"

          strText = sbadmclExecuteCommand(xmldom,0)

           

           

          Manual command line would be:

          sbadmcl.exe -adminuser:admin2 -adminpwd:12345 -command:dumpmachineusers -machine:test -group:* -file:c:\temp\users.txt -expandgroups:true

           

          Faster dump would be.

          (provided you somehow could tell the script what group the machine is in)

          sbadmcl.exe -adminuser:admin2 -adminpwd:12345 -command:dumpmachineusers -machine:test -group:ActualMachineGroup -file:c:\temp\users.txt -expandgroups:true

           

          to sum it up, the more users/machines you have the longer autoD will take to run on each client because it has to search all machine groups for the machine, then expand all the user groups assigned to the machine to determine if the current user is assigned to that machine object.

           

           

          • 2. Re: AutoD - DumpMachineUsers Question

            Thanks for the response.  I ran the sbadmcl command on the client and it took 5 minutes to complete. Running it on EEM also took 5 minutes.  The sbadmcl process on EEM used 20% CPU.  That's a surprising amount of CPU usage. Based on that I think I may have to significantly reduce the number of machines I deploy to at once.  Problem with that is my management is pushing hard to get the remaining 2000 machines encrypted.  Smaller deployment numbers means longer deployment schedule, they're not going to be happy about that.  I guess the only thing that can be done to improve DumpMachineUsers performance is to upgrade the EEM hardware?  Currently I'm running EEM on a VM with 4 CPU and 4 GB of memory W/SAN storage.  May have to move to a physical machine. Could slow DumpMachineUsers cause AutoD to fail assigning the currently logged on user to the machine?  This happens quite a bit.  For example, I deployed 300 machines recently and 86 users failed to get assigned.

            • 3. Re: AutoD - DumpMachineUsers Question

              The same commands performed with large database (~200K objects) complete in less than 3 and 5 seconds respectively. There must be something wrong with your setup (index/cache is not working properly). While using VM server is not recommended, I suspect your problem being somewhere else. Follow Best Practise Guide and troubleshoot index/cache issues. Maybe during those 5 minutes cache is being spontantenously rebuilt - which has very bad implications.

              Check your cache names.xxx files - their size and timestamps when you wait those 5 minutes.

              1 of 1 people found this helpful
              • 4. Re: AutoD - DumpMachineUsers Question

                Besides performance issues what do you mean by" very bad implications?"

                • 5. Re: AutoD - DumpMachineUsers Question

                  As I recall, we had once this situation and some objects were not properly created (ended up as orphans or worse - had to be cleaned up via ceanupgroup commands). Lots of automation tasks failed at that time - and that means a lot of manual corrections have been made.

                  1 of 1 people found this helpful
                  • 6. Re: AutoD - DumpMachineUsers Question

                    Yeah, I'm hoping that doesn't happen.  BTW, I'm trying to run the toastcache.bat but it doesn't seem to be working.  I contacted our rep today and he told me to use the following syntax:  sbadmcl -toastcahe -adminuser:user -adminpwd:pwd

                     

                    All that's happenin is a connection is made to the server then it ends.  Is this the right syntax?

                    • 7. Re: AutoD - DumpMachineUsers Question

                      "sbadmcl -toastcahe ...." ?

                      Of course not!

                       

                      Try to perform this at database server as part of scheduled down time (maintenance):

                      • stop WebHelpdesk Recovery service(s), wait some time
                      • stop Database server service(s), wait some time
                      • delete cache files (names.*), wait some time
                      • sbadmcl -command:getcounts on local database. This rebuilds cache. Wait some time
                      • start Database server service(s), wait some time
                      • start WebHelpdesk Recovery service(s)
                      1 of 1 people found this helpful
                      • 8. Re: AutoD - DumpMachineUsers Question

                        toastcache.vbs /  .bat is a script - you can get it off the tools CD I think, if not platinum support.

                        • 9. Re: AutoD - DumpMachineUsers Question

                          Thanks alot Peter.  I got everything except the delete cache files (names.*) part.   So when I go to my 00000001 and 00000002 folders and drill down through the sub folders I finally get to xxxxxx.DAT files.  Are these the files I'm supposed to delete?  Sorry, still pretty new to this stuff.  Hard to tell right :>)

                           

                          I see a bunch of folders with .WPE extensions also, they're empty though.  Are these deleted machines and users objects?  I thought I read that somewhere before.  Can these folders be deleted as well?

                          1 2 3 4 Previous Next