Hi, I wonder if anybody knows what situations would cause the remote servers in a pool to be used ?
For example if we had one virtual server containing 4 local real servers and 1 remote real server, would connections be directed to the remote server only when the 4 real servers stopped responding to http requests, or would connections be directed there when the max connections were reached on the local servers ? Any help would be appreciated. Regards Andy -------------- next part -------------- An HTML attachment was scrubbed... URL: https://puck.nether.net/pipermail/foundry-nsp/attachments/20041101/b71da0fd/attachment.html From [EMAIL PROTECTED] Tue Nov 2 13:13:40 2004 From: [EMAIL PROTECTED] (David Freedman) Date: Tue Nov 2 13:20:33 2004 Subject: [f-nsp] Multipath iBGP question Message-ID: <[EMAIL PROTECTED]> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, I know that ECMP (Equal cost multipath) for iBGP is a bad idea, so this is only a "lab" question (i.e this is not in production!) here is the question: (apologies if this ASCII art is not formatted correctly) [Router A]------------------[Router B] ~ | | ~ | | [Access Router A] [Access Router B] ~ | | ~ | | [Customer Prefix] [Customer Prefix] ~ 10.0.0.1/32 10.0.0.1/32 Both Router A and B are BI4K MGR4 Ironcore devices with the following configured: ! ip load-sharing 4 ! router bgp maximum paths 4 ! Access Router A and B connected to them are belonging to $other_vendor, redistributing their statics for the customer prefix (10.0.0.1/32) into iBGP. How do the Bigiron RouterA/B know what the correct next-hop is for the customer prefix using multipath? In the lab, traffic arriving on Router A destined for the customer prefix (10.0.0.1) is being sent to Router B Upon arriving at Router B its then being sent back to Router A sometimes!!! Both devices are running 07.06.01B and have also been running 07.06.05 I would have assumed that the process would have instructed the CAMs to have been built in such a way as specifically NOT to send traffic back in the direction from which it came. Is this not the case? The implementation of multipath seems to work well on the $other_vendor kit.... Thanks in advance. - -- David Freedman Network Engineering Department Claranet UK Limited http://www.clara.net -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.3 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFBh83+tFWeqpgEZrIRAtJUAKCsnLdn1PaYblSiOj52s71GsHOiTQCgmjJ0 Bq9zSfQlMtsVyURNPqzSgTA= =B9eo -----END PGP SIGNATURE-----
