Hello,
we implemented "virtual hosting" via kemp lb, i.e. we have one public ip with serveral domain-names and route the traffic via kemp (6.0.44) to 2 different backend servers which have different webservices on different ports.
i.e.
domain-a -> vip 1.2.3.4 -> backendserver1:portX (weight 1000) backendserver2:portX (weight 1)
domain-b -> vip 1.2.3.4 -> backendserver1:porty (weight 1000) backendserver2:portY (weight 1)
and so on.
we added content switching rules for the different domains (header check for "Host:" (virtual host) (e.g. ^domain-a$ regex) and observe strange behaviour. we first thought that a proxy server involved was the culprit, but in a long debug session we found that kemp lb conent-switching goes wrong, i.e. the request for domain-a is "switched" to the backend for domain-b.
for our curiousity, we can NOT reproduce this behaviour from different normal IPs, but whenever the regest originates from the companies central proxy server, we opserve that the kemp is doing wrong - i.e. the request going into the kemp is correct, but the request going from the kemp to the backend is routed to the wrong destination. it looks the content rule engine is on error.
are there any known issues with content switching or did anybody observe something similar ?
regards
Roland
we implemented "virtual hosting" via kemp lb, i.e. we have one public ip with serveral domain-names and route the traffic via kemp (6.0.44) to 2 different backend servers which have different webservices on different ports.
i.e.
domain-a -> vip 1.2.3.4 -> backendserver1:portX (weight 1000) backendserver2:portX (weight 1)
domain-b -> vip 1.2.3.4 -> backendserver1:porty (weight 1000) backendserver2:portY (weight 1)
and so on.
we added content switching rules for the different domains (header check for "Host:" (virtual host) (e.g. ^domain-a$ regex) and observe strange behaviour. we first thought that a proxy server involved was the culprit, but in a long debug session we found that kemp lb conent-switching goes wrong, i.e. the request for domain-a is "switched" to the backend for domain-b.
for our curiousity, we can NOT reproduce this behaviour from different normal IPs, but whenever the regest originates from the companies central proxy server, we opserve that the kemp is doing wrong - i.e. the request going into the kemp is correct, but the request going from the kemp to the backend is routed to the wrong destination. it looks the content rule engine is on error.
are there any known issues with content switching or did anybody observe something similar ?
regards
Roland