Forum Discussion
pelle_71294
Nimbostratus
NimbostratusNode down and 3-way handshake
Hi everyone,
I got a problem with the way our big-ip handles a node down.
First some background, we're running BIG-IP 9.3.1 Build 37.1
The virtual server is loadbalancing memcached to a single node in a pool. It uses SNAT.
When the node goes down there's still a 3-way handshake before a reset is sent. I have verified (with setting up a test-vip with a failing realserver) that the actual server is not the one responding with a 3-way handshake.
Sending a SYN to the realserver will directly respond with a reset. But going through the loadbalancer it results in the 3-way handshake.
Is there anyway I can send the reset directly without the loadbalancer doing a 3-way handshake?
Been scouting through manuals and testing setting "Action On Service Down" to reject but it's still a 3-way handshake performed.
Any help or suggestions is most welcome!
.pelle
hoolioCirrostratus
Hi Pelle,
The quick answer is no. For details you can check this post:
http://devcentral.f5.com/Default.aspx?tabid=53&view=topic&postid=55930&ptarget=55932 (Click here)
Aaron- pelle_71294Hi Hoolio,
I'll test some with the iRule you posted and see if that can help us out.
Thank you for your quick response and help.
.pelle - hoolioWhen using a standard TCP virtual server, you won't be able to avoid the initial three way handshake. The iRule in the other post just ensures an immediate reset is sent after the initial TCP handshake if the pool is down.
Aaron