Forum Discussion
hc_andy_35682
Nimbostratus
NimbostratusDifference between ICMP and GATEWAY_ICMP
Hi All,
This is a pretty dumb question, but what's the difference between using the ICMP and GATEWAY_ICMP health monitors.
I've read up on it but still not 100% sure what...
Minn_62043
Cirrostratus
CirrostratusGateway ICMP has a special purpose.
https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/ltm_configuration_guide_10_0_0/ltm_appendixa_monitor_types.html
A Gateway ICMP type of monitor has a special purpose. You use this monitor for a pool that implements gateway failsafe for high availability.
A Gateway ICMP monitor functions the same way as an ICMP monitor, except that you can apply a Gateway ICMP monitor to a pool member. (Remember that you can apply an ICMP monitor to a node only and not a pool member.)
MOHIT_125417Altostratus
Good one..- Yuval_BA_249670
I still fail to understand the difference. Why isn't it possible to use the regular icmp monitor on a pool?
IanBEmployee
Pool members are a combination of address and port, whereas nodes are just an address. A pool member describes a TCP or UDP service, and it makes no sense to ping a service, since ICMP has no concept of ports, so this is why we don't allow the ICMP monitor to be applied to a pool member.
The gateway_icmp monitor was intended to be used in transparent mode, where it would be configured to ping a specific address (different from the node), and in that scenario, it makes sense to apply it to a pool member.
Of course, if you don't configure it as transparent, then it will ping the pool member's host, and effectively perform the same check as an ICMP monitor applied to the node. It should be understood that you're not checking the service - you're just pinging the host that the service is running on, and if the webserver went down, a gateway icmp monitor would continue to report the host as being up.
- clutch2k3_27392
While this statement in and of itself is true: "it makes no sense to ping a service, since ICMP has no concept of ports", it doesn't mean f5 should arbitrarily decide to not include the icmp monitor for normal loadbalancing pools. I'm sure that the f5 developers are aware that users are "hacking" their pools by applying the gateway_icmp monitor rather than configuring each individual node in the pool with an icmp monitor. Maybe the f5 dev team should listen to their users instead of offering a straw man argument?