7 Replies Latest reply on Sep 16, 2011 10:36 AM by THogland

    LDMS9 overloading Oracle 11g

    THogland Apprentice

      Since upgrading to LDMS9 sp2, we've had problems with the inventory service dropping offline. Our Oracle DBA was noticing the database listener would stop responding, which seemed to cause the service problems, and has been looking at Oracle. This morning, after rolling out patches over the weekend, I noticed that the inventory data is badly out of date on many machines, including mine. (Example - system does daily inventory scans, but machines are showing nothing since late July or early August.) I had clened out the ErrorScan folder 10 days ago, and in the last 10 days I have ~7300 failures in there (about 3/4 of all scans, if my ~1000 machines are scanning daily for the last 10 days.) Asked the DBA, and he had finally determined that LANDesk is overloading the listeners with the number of concurrent database sessions trying to connect to Oracle, and asked that I limit LDMS to 120 concurrent connections.


      Is there a way to do something like this? I can't find anything in the community about it...


      This isn't a huge system, either - one core, a dozen total users, about 1000 devices.

        • 1. Re: LDMS9 overloading Oracle 11g

          Are you sure with the Overloading?

          You may check your Agent Options, how often you make Scans and so on,

          to reduce this.

          I would suggest you to open a case for this.


          Best Regards


          • 2. Re: LDMS9 overloading Oracle 11g
            THogland Apprentice

            Yes, definitely overloaded - the Oracle DBA can see hundreds of open sessions across all four servers, and I'm still getting inventory failures. Agents were set to scan daily, but the real-time inventory scanners were enabled. I've pushed out a new agent deleting those, and we'll see if it reduces the load.


            Issuing 'iisreset' on the core clears out most of the sessions. I'm currently digging for some IIS tuning docs for v9, but found nothing so far...


            A typical event log entry:


            Event Type:     Error
            Event Source:     LANDesk Inventory Server
            Event Category:     None
            Event ID:     4100
            Date:          9/14/2011
            Time:          11:17:48 AM
            User:          N/A
            Computer:     DNRATW-LD01
            Database exception: SCA1732.SCN, LDInv.dll
            Item cannot be found in the collection corresponding to the requested name or ordinal.
            Update Table:TCP Column:ADDRESS PK:, Thread ID: 8284.


            That's the most common error, although I also see a description of:

            Clearing column INSTALLDATE from APPSOFTWARESUITES

            7428, Thread ID: 7428.

            • 3. Re: LDMS9 overloading Oracle 11g

              Are you using the latest MCP Patches (0519?

              I remember that in one (Core) of them, was a fix for Oracle Database with updating PK.

              You may ask the Support for it.


              >>Feb 2011 MCP Datamart Creates invalid PK column leading to invalid triggers for Oracle users.


              Best Regards


              • 4. Re: LDMS9 overloading Oracle 11g
                BigB SupportEmployee

                Limiting the # of concurrent sessions will seriously cripple your LANDesk functionality. If you are managing 1000 nodes, then you should have as many as 200+ concurrent sessions running. LANDesk requires a database with close to dedicated database funcitionality.

                • 5. Re: LDMS9 overloading Oracle 11g
                  THogland Apprentice

                  I've currently got 200+ connections allowed on each of the four nodes - 800 total. We can push that closer to 300 if needed...

                  • 6. Re: LDMS9 overloading Oracle 11g
                    MarXtar ITSMMVPGroup

                    What do you mean by having four nodes? Have you setup multiple inventory services? For a 1,000 node environment this sound like a bit of overkill if you have.Not necessarily related to your issue, but sounded odd.


                    Definitely look into the patch since it sounds like what you are encountering, and having constant errors certainly won't be helping your performance.


                    Mark McGinn

                    MarXtar Ltd



                    The One-Stop Shop for LANDesk Enhancements

                    - Wake-On-WAN - Distributed Wake-On-LAN, Scheduled Power Down, and SWDist Sequencing

                    - Inventory Archiving/Change & Root Cause Analysis

                    - Self-Service Password Reset

                    - PC Power Management

                    - Help/Service Desk Integration

                    - LANDesk Data Manipulation/Integration/Bar Coding

                    - LANDesk Executive Reporting Pack

                    - Network Asset Discovery and Management

                    - LANDesk Software Store

                    • 7. Re: LDMS9 overloading Oracle 11g
                      THogland Apprentice

                      Sorry for the lack of clarity - I have two threads and a problem report open...


                      The core is a VMWare VM, WIndows Server 2003, 2 cores, 4GB RAM. Our Oracle is 11gR2 RAC - four front-end servers performing load balancing into a large multi-disk array. LANDesk was set to use only one node until recently, when it's 1000+ concurrent connections effectively killed that one node (refusing logins, etc.). The DBA set it to 120 connections per node, and allowed it to use all four nodes at once, which (according to LD) should have been sufficient; when we still had errors he increased it to 250 connections per node, or 1000 concurrent connections allowed.


                      LD Tech Support is digging into why this is happening, and why the errors seemed to snowball lately. I'm not going to apply any patches until I hear back from them. At least not yet - my OSD imaging is still working, and scripting seems to work, but anything query-related is broken, since inventory is horribly wrong and getting worse.