11 Replies Latest reply on Aug 2, 2011 3:18 AM by Ian

    Peer download freezes after transferring exactly 59Mb of a file

    Rookie

      Hello,

       

      has anybody else encountered a file size limitation on deploying software? I now have two tasks, both stuck-on-active because the peerdownload system freezes at 59,0 MB (61.888.610 bytes). I don't know how to force it to 'download from core only'.

       

      A little background: I am running two versions of landesk on the same subnet, some old systems are completely 8.7sp3 and talking to the old server, while a dozen or so are 9.0sp2 and talking to a new server (which should become the only one eventually, but it has still not made it out of testing). Usually the mixed environment works fine, but today I am hitting a snag.

       

      The 8.7sp3 systems had no problem whatsoever with a batchfiletask (1.4Gb in about 70 files, standard push without policy) running off of the old environment, installing just fine.

       

      About two hours later, the exact same task is being deployed to the 9.0sp2 systems, from the corresponding core-server with its own storage (identical path and filenames as the 8.7sp3 deployment package, just on a different server). From the logfiles I see the 9.0sp2 clients downloading files from the 8.7sp3 clients, which should be fine - except for the full freeze at 59.0Mb. No errors, task remains active, can't cancel it, can't restart it, can't start other tasks, unless I logon to the client, kill sdclient.exe manually, and restart the machine. Not nice.

       

      Question1 - can I force the clients to download from core only, bypassing the entire peer-download system?

       

      Question2 - shouldn't clients be restricted to only download from peers at the same coreserver? Or, differently put, can I somehow convince peer87 that a file downloaded from landeskserver87 should not be offered to a peer 90 that looks for files on landeskserver90?

       

      Regards,

        • 1. Re: Peer download freezes after transferring exactly 59Mb of a file
          Expert

          What happens after the download? Is something executed? Is a file being locked because its in use?

           

          The setting you are looking for is in the network section of your Delivery method. Have a look at the pic attached (from 9.0 sp2).

           

          FYI, agents use the core server name and a cert to identify which core to communicate with, which is why it works ok with a mix, but I would assume that at some point you may hit some snags.

          • 2. Re: Peer download freezes after transferring exactly 59Mb of a file
            Expert

            An addition to question 2. You could also use Preferred servers (file servers local to the clients designated by subnet). The clients will go to the preferred server (if available) and get what they need from there.

            1 of 1 people found this helpful
            • 3. Re: Peer download freezes after transferring exactly 59Mb of a file
              Rookie

              'After download' never happens - the download just freezes with a nice @@partial@@ of 59Mb. On both ends I see the connection alive in tcpview, but no data is transferred anymore, but since there is no error, sdclient does not recognize it as an errorcondition and just waits for more data (which never comes).

               

              Didn't check for locked files, since the transfer is already halfway I don't think they are.

               

              Both cores have a different certificate for communication, but I have added the old certificate to the new core to do some testing with pushing new agents to old clients - I'll remove the certificate, and see if that also stops the peers from communicating between eachother.

               

              As for the deployment options - been there, done that :-( Multicast is too slow (since it slows down to the slowest client - only one busy machines), 'download from source' is my default but does *not* disable the peer-download system (despite it being implied). Haven't tried 'run from source' yet - I assumed that would only work for single-file packages?

               

              I'll look at the preferred servers, if it works as I'd expect it could just do the trick.

              • 4. Re: Peer download freezes after transferring exactly 59Mb of a file
                Expert

                The following is my opinion, so don't take it a best practice or hardend rule.

                 

                LANDesk uses something similar to MS's BITS to trickle the files local before deployment when using Multicast or Download from source.

                 

                Multicast is ok in envoronments where you have limited resources and a slow wan link like a dialup line to a remote office (think of a logging company with office trailers in the bush, and yes, I actually had this scenario). One agent trickles down the package and becomes a "Preferred server" for the rest.

                 

                Run from source works ok, but can add network traffic and slow the overall speed of the deployment.

                 

                Download from source is the one that I have had probably the best result with because once all the files are local, then the local agent scheduler performs the install and after 14 days (default setting) the sdmcache is cleaned up.

                 

                Preferred servers are a good idea where WAN links or remote subnets are used or you have only 1 core and have multiple of subnets and a large amount of clients per subnet.

                 

                Again, just my opinion and understanding of how it all works.

                • 5. Re: Peer download freezes after transferring exactly 59Mb of a file
                  mrspike SSMMVPGroup

                  I have found that when I see @@filename.xxx@@ in the cache and it fails is that one of the peers has a corrupt copy of the patch or file.

                   

                   

                  I created a "Manage Script" that I can push out to all of my system or just systems on the same subnet, that will nuke the cache folder.  it workd well for us.

                   

                   

                  [MACHINES]
                  REMDEL01=%ldms_Client_dir%\sdmcache\*.*

                  • 6. Re: Peer download freezes after transferring exactly 59Mb of a file
                    ProServices

                    Hi,

                    Could you try it again with a distribution method using 100% bandwidth please. Also, does changing the % Bandwidth change the point it gets stuck at?

                    Regards

                    Ian

                    • 7. Re: Peer download freezes after transferring exactly 59Mb of a file
                      Rookie

                      Ian,

                      I've tried various deployment types, they all failed at exactly the same point, but only when 9.0 clients tried to download from 8.7 peers. Downloading from identical-versioned peers or core - no problem.

                      Haven't seen the problem since, because I took precautions to not deploy large files simultaneous on both versions or clear the old caches before deploying to 9.0 clients. Hopefully will not see it again because I'm now rolling 9.0sp2 to all workstations, which removes the entire mixed-version problem anyway.

                       

                      Regards,

                      Johan

                      • 8. Re: Peer download freezes after transferring exactly 59Mb of a file
                        ProServices

                        Hi,

                        I saw this for the first time a couple of months ago,on a very large site. Basically, when the 9 agent talks to an 8 peer to do the peer to peer copy it all starts off OK. However, 9 has some LAN bandwidth control options for the download that the 8 agent can not understand. So, if the LAN bandwidth is not set to 100%, the 9 agent downloads a bit and then sleeps a while. The 8 agent does not understand what is happening and is unable to continue. The 9 agent never times out!

                         

                        So, the good news - there is a post SP patch for 8 (NOT 9!) which should fix this issue. Ask support. However, there is also a work arround, set the LAN bandwidth to 100% ( you cant set it to 0% which would turn off the peer to peer function because the GUI wont let you go below 1%. The patch is a good solution if you feel confident that you can hit 100% of your machines. Due to the nature of the problem, you only need one unpatched device per subnet to cause you random on-going problems. If you dont think you can hit 100% of your devices I suggest to go with your current work arround, or the 100% LAN bandwidth option.

                         

                        Regards

                        Ian

                        • 9. Re: Peer download freezes after transferring exactly 59Mb of a file
                          Catalysttgj Expert

                          Sounds like we're experiencing the same problem between 9.02 and 8.7 sp5 agent peers. We'll be testing the patch and probably the workaround tomorrow.

                           

                          Note: We see the problem with vulscan downloading large patches.

                          • 10. Re: Peer download freezes after transferring exactly 59Mb of a file
                            Catalysttgj Expert

                            Just wanted to post another update about this issue. There is NO patch to fix this problem on 8. This has been confirmed with our TAM, a PSE, and engineering.

                            We're having another conference call next week on this issue with LANDesk to see if they might decide to create a patch for it, but it would be a 9 side patch, not an 8 patch, but just so anyone who might read this in the future, there is no patch at this time for this issue.

                            • 11. Re: Peer download freezes after transferring exactly 59Mb of a file
                              ProServices

                              Hi,

                              Just to correct my earlier post.. the patch I was told about seems not to exist. The workarround is to use 100% bandwidth for peer-to-peer transfer (which can be set in the GUI) or use 0% (which is very hard to set in the DB AgentBehaviour and I would not reccomend trying).

                              Regards

                              Ian