Forum Discussion

brianokelly_119's avatar
Icon for Nimbostratus rankNimbostratus
Feb 22, 2011

LTM Health Monitor Issue

Hi all, probably the best place to post this. We have a pair of 8900s working in Active/Strandby mode. They are running multiple partitions with separate route domains. The F5 sits in the traffic path with public addressing used for the VIPs and RFC1918 addressing internally. In one partition, we have a web application VIP distributing towards multiple Oracle Web Logic servers. Our problem is at the most basic level, the ICMP node health checks are failing as the source IP address of the health checks is being set to the public address of the F5 but is being routed via the private connected Internal interface. I have never come across this issue before, I assumed the routing table would resolve it. To explain better here are some example values:



Self IP External:


Self IP Internal:


Node IP Addresses to load balance against:,,



The routing table is: via


Default Route via



The pings towards are being sent via but the source IP address is set to I have never seen a mechanism to set the source IP address of health checks. Can anyone think of anything that could be set incorrectly?


5 Replies

  • Hi Brian,



    I thought the monitoring connections would follow the TMM routing as well. There isn't a config option for explicitly setting the source address or VLAN for monitor connections. I wonder if the route domain config is affecting this.



    Are the pool members in the default route domain or a custom one? Is the default route for the default route domain?



  • Hi Aaron,



    All routes are configured in this partitions route domain. All Self IPs are also. They should all be completely isolated.



  • I'm not sure what would cause this? Can you open a case with F5 Support to have them review your configuration? If/when you figure this out, could you reply back here for future reference?



    Thanks, Aaron
  • Hello,



    did you get a reply to this perchance? Did you encounter the issue in 10.2 code?



    I've seen this problem once on this code version and I understand it was due to be fixed in the 10.2.1 release.



    Sorry for the late response :S