CSM: predictor hash url vs predictor hash address source

Unanswered Question

I have a CSM set up as a transparent load-balancer for our web caching appliances (bluecoat).


When using predictor hash url, certain web apps break when the client moves between cache engines. It appears the applications are using IP address as the session identifier and when the URL changes significantly the client is balanced over to another cache engine--breaking the session.


In order to work-around poorly written web applications I have had to resort to source-ip hashing. This has resulted in an uneven load distribution across our caches.


Does anyone else do transparent web cache load-balancing and have similar problems? If you are using url hashing, how did you configure your vservers and serverfarms?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
brispin Fri, 08/31/2007 - 10:43
User Badges:

For URL hash method

url-hash begin-pattern [end-pattern ]

All characters between and include the keywords are included in the hash-value.The CSM looks at the first few bits in the hash-value in order to mapit to a real. If this real is "disabled", the CSM find the nextavailable real in the list.



Verify this bug for predictor hash issue:CSCeh03583

Actions

This Discussion