Jump to content
Welcome to our new Citrix community!

Active-Active GSLB in round robin with persistence


Amar Aditya

Recommended Posts

Hi,

I am working on a GSLB active-active setup, wherein configured lb method is round robin & also has persistence.

Complete setup :-

===

4 back-end servers directly the part of the gslb.

LB method is in round robin.

persistence 5 min

 

Issue :- Now what i am noticing almost all the traffic is being sent to only 1st 2 servers most of the time. though post 5 min of idle time.

 

Also, weight/ratio cannot be deployed- not as its not part of standard build process. 

 

    

Link to comment
Share on other sites

Can you show your gslb runningconfig and confirm you have 4 gslb services bound to 1 gslb vserver?

Which persistence type are you using source ip or other method?

 

How are you testing?

If you test from the NetScaler or via ping/nslookup from a client you should confirm the dns resolution does in fact roundrobin (if persistence is removed).

Plus if you reconnect before the persistence expires, it may continue using persistence.

During a browser based test, it may not as the browser has to actually make a new dns lookup and it may be caching which could skew your results.

 

 

 

Link to comment
Share on other sites

Persistence Type :- Source IP

 For testing - i wrote a simple a  script with 7 min interval querying (Nslookup) the both LDNS & ADNS servers, for such 20 requests.  

o/p :- Most of the time it went to one or the other server  but not distributing across all the services 

 

Link to comment
Share on other sites

 

What is the persistence timeout AND can you show your gslb service bindings to the gslb vserver.

 

With too little traffic it might not actually round robin as connection tables fall back to zero, so there's no reason to load balance in a continuous order, as a new load balancing cycle just needs to be distributed regardless of prior requests. And if all prior connections have timed out, there's nothing to base the next round robin decision on.  If you actually had multiple clients test, you're more likely to see your load distribution.

 

But a simple ping test that starts and stops multiple times should be able to confirm round robin, if all services are up and all gslb communication is functioning.

First test: take persistence off and ping 8-10 times to confirm round robin. If its still not doing round robin, then check the services states and gslb site to site connectivity.  

Once you add in persistence, its better to check with multiple clients to confirm the traffic distribution mechanism; so run your script from 4 different clients to confirm gslb.  

 

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...