<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">2014-02-23 16:18 GMT+04:00 Srinath C <span dir="ltr"><<a href="mailto:srinath.c@gmail.com" target="_blank">srinath.c@gmail.com</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
when the failed node recovers it will remain up and running with no connections established to it since all the clients are having persistent connections to the other node. The load is now skewed and all the load is thrust on one server with the other server lying idle. Are there common solutions to this kind of load sharing problems? Please advice.</blockquote>
</div><br>Stick HAproxy in front of the cluster and make clients connect to it. It can load balance incoming</div><div class="gmail_extra">connections and keep track of the backend (node) states and won't send connections to the backends</div>
<div class="gmail_extra">that are unavailable.<br>-- <br>MK<br><br><a href="http://github.com/michaelklishin" target="_blank">http://github.com/michaelklishin</a><br><a href="http://twitter.com/michaelklishin" target="_blank">http://twitter.com/michaelklishin</a><br>
</div></div>