F5 load balancing problem

Hi Leonard -

 

 

if the web servers at site A is down then traffic will be routed to site B web servers.

 

 

How is that accomplished?

 

 

/deb

 

 

We have a DNS that points to Site A and Site B load balancers. The F5 load balancers at Site A is configured with a higher priority for servers located in Site A.

 

 

Our problem is that some how users connecting at the middle of a session with site A web servers out of a sudden they will be redirected to site B web servers. We pointed out that the problem maybe due to the DNS timeout session.

 

 

However can you include an i rule to make Site A servers always the priority unless the servers went down or what?

You wouldn't want to force traffic to a server LTM thinks is down. You should try to figure out why the traffic is switching and fix that issue instead of implementing a brute force workaround.

 

 

So if I understand correctly, you have something like this set up?

 

 

Site A virtual server

 

...pool member SiteA_Server priority 10

 

...pool member SiteB_Server priority 5

 

 

Site B virtual server

 

...pool member SiteB_Server priority 10

 

...pool member SiteA_Server priority 5

 

 

and you're using standard DNS to resolve the virtual server address?

 

 

If that's the case, there are a couple of things you might want to look into to determine what's going on.

 

 

DNS switching to different virtual server

 

If DNS hands out 2 answers, the client may decide to use either of them fairly indiscriminately, typically choosing one until it fails to respond.

 

 

Regardless of how many answers are returned by DNS, the answers have a defined timeout (TTL) and the client will eventually have to resolve the name again, and there is no guarantee which answer will be sent or used.

 

 

You could see if the same client is hitting both virtuals by running a packet trace on the LTM external interface @ both sites & looking for the same client sourceIP.

 

 

LTM switching to different pool member

 

If the local server appears down to LTM, it will send traffic to the remote server instead. You're probably using a SNAT to ensure the return traffic is routed appropriately, and if so, examination of the server logs would show you which LTM sourced the connection. If no SNAT is in place, you can see if the same client is using both servers in the pool by running a packet trace on the LTM internal interface & looking for the same client sourceIP going to both pool members, then examining the LTM logs to see if the local pool member was marked down mid-session.

 

 

If you see that the monitored server status changed, verify your actual server status & adjust the monitor configuration if LTM is marking your nodes down too aggressively.

 

 

HTH

 

/deb

And I suppose it wouldn't hurt to mention that our GTM product is incredibly useful for managing (and troubleshooting) multi-datacenter issues like this.

 

 

/deb

Thanks deb it is exactly what I am facing now. I supposed the problem is more to DNS switching to different virtual server. I will be introducing the GTM to my customers soon. But however I will need a workaround. Are there any suggestions to prevent the users from losing their session after the DNS TTL is lost?

None I'm aware of.

 

 

GTM gives you the ability to persistently hand out the same answer to the same LDNS, and perhaps other intelligent DNS products offer similar functionality, but you can't persist an LDNS to a specific A record in standard DNS implementations.

 

 

/deb