none
Could not retrieve Enterprise Global Template - Load balancer issue RRS feed

  • Question

  • Hi,

    We have 4 Project Server 2010 servers. The 4 web servers are load balanced by networking team with sticky session configured.

    When we try to connect to the Project Server using MPP 2007 SP2, it fails saying 'Could not retrieve Enterprise Global template'. It works perfect when we point to a specific server by specifying the IP address for server name in the 'hosts' file.

    Earlier we observed some errors in the event viewer related to the SharePoint's internal load balancer for which restarted the 'Project Server Application' on each web server and it got fixed.

    Now, the only entries that we see related to load balancer are as mentioned below as Information (not errors).

    -------------------------------------

    SharePoint Web Services Round Robin Service Load Balancer Event: Initialization
    Process Name: w3wp
    Process ID: 15080
    AppDomain Name: /LM/W3SVC/539065287/ROOT-1-130462463500778047
    AppDomain ID: 2
    Service Application Uri: urn:schemas-microsoft-com:sharepoint:service:ae7c7ee5c09b4e8198bdbb1ecb8c1c1b#authority=urn:uuid:9f626d347784423eb14bde4a1f4d13fc&authority=https://lonms12546:32844/Topology/topology.svc
    Active Endpoints: 4
    Failed Endpoints:0
    Endpoint List:
    http://lonxxx2532:32843/ae7c7ee5c09b4e8198bdbb1ecb8c1c1b/PSI
    http://lonxxx2545:32843/ae7c7ee5c09b4e8198bdbb1ecb8c1c1b/PSI
    http://lonxxx2546:32843/ae7c7ee5c09b4e8198bdbb1ecb8c1c1b/PSI

    http://lonxxx2566:32843/ae7c7ee5c09b4e8198bdbb1ecb8c1c1b/PSI

    --------------------------------------------

    Could the issue be due to network load balancer?

    Could the issue be due to Sticky session configuration on the load balancer.?

    How can we get to the root cause of the issue?

    Which logging category should we set to 'Verbose' that can give us some hint.

    Update: We tried to capture the requests through fiddler and observed that when fiddler is running on the client computer then the connection works perfectly fine even through the load balancer. Probably fiddler is reformatting the SOAP envelop of the web service requests the way it should before sending the request to the server.

    If we do not run fiddler and run some other similar tool (like Charles) then it again gives the issue and the request stucks at /PWA/_vti_bin/psi/winproj.asmx

    We ran Wireshark on the servers and found the following for that web service call:
    [TCP Previous segment not captured] Continuation or non-HTTP traffic.

    Please let me know if someone could provide any hint what can be done next.


    Regards, Amit Gupta



    • Changed type amitgg Wednesday, June 4, 2014 7:15 AM
    • Edited by amitgg Wednesday, June 4, 2014 7:25 AM
    Tuesday, June 3, 2014 5:38 AM

Answers

  • Finally it turned out to be a Load Balancer issue.

    As per network engineers...

    In case the IP packet lenght is less than approx. 700, the loadbalancer forward this to backend server and everything is fine.

    In case this packet is more than 700, loadbalancer ignores the packet, and sending host after several retransmittion attempts gives up and opens new connection.

    So they did some configuration on the load balancer and it got fixed. :)


    Regards, Amit Gupta

    • Marked as answer by amitgg Thursday, November 13, 2014 10:46 AM
    Thursday, November 13, 2014 10:46 AM

All replies

  • There are several ways to configure your load balancer.   I would suggest that you work with the network engineer, the load balancer vendor and your project administrator to resolve this issue. 

    Basically you need URL to be resolved correctly.  Also, I don't believe PS2007 did a good job handling load balancing, so you may need to bring someone in good with IIS and see they can tweek IIS to manage the cache better.

    As I go back and look at your analysis, I think you should probably look at upgrading to Project Server 2013.  They made some improvement in load balancing and the management of distributive cache.

    I assume you have 4 WFE because you have thousands of project users.  Roughly how many  you have?  Over 1000, over 5000

    Have you tried to see if using two load balancing work?  How about just one front end.  I often see companies scaling SharePoint and Project server to extremes. 


    Michael Wharton, MVP, MBA, PMP, MCT, MCTS, MCSD, MCSE+I, MCDBA
    Website http://www.WhartonComputer.com
    Blog http://MyProjectExpert.com contains my field notes and SQL queries

    Thursday, June 12, 2014 3:31 AM
    Moderator
  • Finally it turned out to be a Load Balancer issue.

    As per network engineers...

    In case the IP packet lenght is less than approx. 700, the loadbalancer forward this to backend server and everything is fine.

    In case this packet is more than 700, loadbalancer ignores the packet, and sending host after several retransmittion attempts gives up and opens new connection.

    So they did some configuration on the load balancer and it got fixed. :)


    Regards, Amit Gupta

    • Marked as answer by amitgg Thursday, November 13, 2014 10:46 AM
    Thursday, November 13, 2014 10:46 AM