|
Hi,
Does the Linux Virtual Server setup take into consideration the load of
the node before sending requests to it. For example if a node was doing
an intensive backup would the load balancer choose a different node
based on its server load?
There is an active and a passive weighting framework. The passive one is called
wlc and is a builtin scheduler within LVS which takes the amount of processed
connection into considerations. So if a server is overloaded due to a backup
naturally it won't be able to process as many requests as the other nodes. Thus
the amount of connections handled by this server will dynamically reduce over
time and this shows up in the amount of active and passive connections.
The active weighting framework can be done with an additional software called
feedbackd [1]. It will report back the load of the RS to the director which in
turn can twiddle with the weights and thus achieve a faster correction of load
imbalance.
[1] http://www.redfishsoftware.com.au/projects/feedbackd/
HTH and best regards,
Roberto Nibali, ratz
--
echo '[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc
| 