Forum Discussion

Stanley_87566's avatar
Stanley_87566
Icon for Nimbostratus rankNimbostratus
Jan 06, 2010

Packet reset when call Virtual Server

Hi All,

 

 

I found a performance problem when an application call a Virtual Server on my BIPIP 3400.

 

 

The problem is some packets will be reset during 3 ways handshake and before load-balancing. Even though some request luckily successfully load balance to the node (web server). The transaction need time is very long. So that transaction timeout caused. However, when that application call the Node (web server) IP address directly, all problem disappeared.

 

 

I'm using "BIG-IP 9.4.5 Build 1086.1 Hotfix HF2" with layer 2 network configuration (External and internal interfaces are same segment and both external and internal interface connecting to 100baseTX full). Therefore, the problem should not because of other network equipment when I tried to bypass the virtual server and the node real IP address.

 

 

Is any have idea or experience it?

 

 

Many thanks!
application delivery
design
performance
testing
  • The_Bhattman's avatar
    The_Bhattman
    Icon for Nimbostratus rankNimbostratus
    Jan 06, 2010
    Have you looked on the ask.f5.com site.

     

     

    Here is search that might reveal the issue for you:

     

    https://support.f5.com/kb/en-us/search_results.advanced.html?productList=big-ip_ltm&versionList=9_4_5&searchType=advanced&query=reset&submit_form=Search&product=big-ip_ltm&documentType=kb%3Aknownissue&productVersion=9_4_5&dateFilter=&num=10&advSyntax=

     

     

    Bhattman

     

  • Stanley_87566's avatar
    Stanley_87566
    Icon for Nimbostratus rankNimbostratus
    Jan 07, 2010
    simular is SOL8867. But it seems not. But I found a PVA acceleator is "FULL" on the default fastL4 profile on 3400. Is it the root cause? Since I cannot find this PVA value on my another BIPIP 1500 environment.

     

     

    Thanks
  • Stanley_87566's avatar
    Stanley_87566
    Icon for Nimbostratus rankNimbostratus
    Jan 07, 2010
    I can simulate the case, but I cannot find the root case and cannot find the solution. Can anyone help? Below is my environment load balancing design

     

     

    Basically, I've 4 servers, let say A, B, C & D assigned to Virtual Server. But B only will backup A when A is down and D only will backup C when Server C is down instead of simple round robbin load balance to all 4 server.

     

     

    So, I've created Pool-1 include Server A & B, but A priority group is higher than B. And I also create Pool-2 include Server C & D, and C priority group will higher than D.

     

     

    But only Pool-3 included all A, B, C & D servers assigned to my Virtual Server. Because of B only backup A and D only backup C. But I need apply an iRule (sample below) to handle B backup A only and D backup C design.

     

     

    LB_Failed {

     

    If A down;

     

    perform reselect Pool-1;

     

    elseif B down;

     

    perform reselect Pool-2;

     

    elseif C down;

     

    perform reselect Pool-1;

     

    elseif D down;

     

    perform reselect Pool-2;

     

    }

     

     

    All the above config. is suggested by F5 and well tested on another environment. However, many packet reset found suddently. But this error disappeared after I used a simple round robbin load balance traffic to all 4 servers or the caller directly make request to the server real IP address. Therefore, the problem should not because of physically connection problem.

     

     

    Many thanks