Does CARP include the node's port in the hash?

Thanks for your reply. I had read the document, which says the hash is computed from each character of the MemberName, and although there is no definition of MemberName in the document, it does define the Name entry in the Member table as the FQDN, rather than the IP and port, and would produce an identical hash for two entries with the same FQDN and differing ports. However, the F5 implementation only has the pool to use as the member table, which means they will use the IP address, but not necessarily the port.

That said, I did a test after my post yesterday, and it does seem to include the port in the hash.

Of course, the primary reason for doing this was to try and put some bias towards some of the proxies. It would be nice if the ratio field in the pool member table could be incorporated into the hash calculation.

Thanks for the information Stephen. It's not the best document and is full of spelling mistakes plus it's expired so it probably wasn't the best place to point you to. Worth noting from the TMOS release notes: "...the hash persistence profile can now perform a pool member selection using a stateless hash algorithm based on the Cache Array Routing Protocol..." which would suggest that F5's implementation is all their own but still implements the core feature. That being the case, it's possible they do consider the port and also that Pool Member ratio could be built in to the calculation.