Forum Discussion

Jaral_369195's avatar
Jaral_369195
Icon for Nimbostratus rankNimbostratus
Aug 07, 2018

Load balancing issues

Hi Everyone, hope you all are doing great!! May i have your kind support to know what we can do or what would be the reason which is causing below issues for one of the application hosted at the site.

 

Issue: Load balancing is not happening across all the servers/Pool members.

 

Brief about Application: I have a web application (https://xyz.jpd.com/abcportal/) hosted on 4 servers, Ser1,Ser2,Ser3,Ser4. All the servers are having same hardware configuration.

 

LB algorithm for pool members: Round Robin

 

Persistence Profile on VIP: Source_addr_mirror_900

 

Source address translation: Auto Map

 

How users are accessing to the application: There are two type of topologies which we have found so far by which users are accessing the application. -Type1: Users are directly accessing the website connected -Type2: Users are first doing RDP to ACP servers and the from ACP sever they are accessing the application.

 

What action taken so far to resolve the issues: Due to persistence users were using already established TCP connections, due to this Ser1 or Ser2 are having almost all the connections and sometimes Ser3 and Ser4, whereas Ser1 and Ser2 were sitting idle, to overcome from this situation and get rid of this, We changed persistence profile from Source_addr_mirror_900 to NONE, Due to this users those where accessing website from Type2( through ACP server) connectivity lost access to the website, Page stopped to open.

 

We had a T-shoot call with ACP server team to have them check any configuration/setting on their server which is stopping users to access the website when persistence is disable on F5 LB but we found no configuration/setting related to persistence on their side as ACP server behave similar to citrix server/AWS.

 

Now, the next and final suspect which we are thinking is get connected with application owner and get setting’s/configuration check on their side to see if they have anything set on their side related to persistence.

 

May I know your what could be the reason of not getting user’s connected to application when persistence is set to none of LB and also would like to know what we can ask application team to check in the application.

 

Please note: Request coming from ACP servers with Only three IP address ACP1,ACP2 and ACP3, since all users are connecting to application through RDP.

 

devops
LTM
  • Sukesh123456's avatar
    Sukesh123456
    Icon for Altostratus rankAltostratus
    Aug 07, 2018

    Hi Jaral,

     

    Can you check if the users coming via RDP session are able to make it to the F5? Please use tmsh show sys connection | grep x.x.x.x command to see that.

     

    Also use this command once before and after disabling the persistency

     

    I suspect that when persistency was enable couple of servers would be responding which were capable.

     

    So lets say out of the 4 servers 2 were capable and now they might not be getting the request.

     

    Thanks!

     

  • JG's avatar
    JG
    Icon for Cumulonimbus rankCumulonimbus
    Aug 08, 2018

    Try to configure a cookie persistence profile for your website, use "Least Connections (member)" as the load balancing method, and enable the use of a "OneConnect" profile.

     

  • Jaral_369195's avatar
    Jaral_369195
    Icon for Nimbostratus rankNimbostratus
    Aug 08, 2018

    Hi Sukesh,

     

    yes, i can see there are only 2 source address are hitting to LB and both are ACP server IP address, Since users are doing RDP to ACP server and then from ACP server they are opening the application, that's why LB is not doing balancing the load to another servers and sending every request to ser1 and ser2 from source as ACP server, Moreover connections are remain on Ser1 and Ser2 due to persistence.

     

    Thanks very much for your suggestions but here i expect more from you, as i have some questions in my mind.