This is not bad, but I would need to add to my scripts check about the
actual load (overhead).
I would prefer if Nginx would be able to receive message from "dynamic
servers".
Benefits would be:
* web application could manipulate how server should be loaded
* any script could exclude the server from LB. for example if I would
run script to get new code from SVN, I would first inform nginx to stop
to use the server and after all code is refreshed I would put myself
back to LB
* any LB strategy could be easy implemented
It would be also great, if there would be a way how to confiure Nginx to
stop to use the server for some time (for example 10 sec) or infinite.
Nobody else would need such functionality ? Or has somebody better idea
for optimal load balancing ?
Manlio Perillo wrote:
Martin Minka ha scritto:
Hello,
is there a way or is there plan to automatically change weight for
server in upstream depending on actual system load ?
I have different HW on 7 apache web servers and it is nightmare to
optimal set the weights.
Not sure, but one of your backend can simply return a 503 response code
when the system load is high.
nginx, then, will proxy the request to another backend:
http://wiki.codemongers.com/NginxHttpProxyModule#proxy_next_upstream
4 messages in ru.sysoev.nginxRe: load balancing depending on actua...
