On Tue, Sep 16, 2008 at 1:09 AM, Rupert Fiasco <rufiasco#gmail.com> wrote:
> We have upgraded all machines to 18.104.22.168 but are *still* seeing a
> mongrel being given more than 1 request at a time. This is not a
> flagrant issue across the board (like it was prior to this latest
> version), but only occurs in say 1/8 mongrel.
We have been running 22.214.171.124 all day with perfect maxconn behaviour. (Thanks, Willy!)
Do you see the problem on the HAProxy status page, or in the Mongrel "ps" status? It's my impression that HAProxy health checks ignore the maxconn setting. So we frequently see this "ps" output:
origo 25720 5.4 3.6 396940 300964 ? Rl Sep15 5:34 mongrel_rails [10006/2/1369]: handling [...]: GET /-/test/app
(The middle "2" is the number of outstanding requests being processed by Mongrel.)
In our case this is acceptable behaviour. In the worst case, if Mongrel does not respond to the health check quickly enough, the backend will be marked as down for a short while.
Alexander. Received on 2008/09/16 01:28
This archive was generated by hypermail 2.2.0 : 2008/09/16 01:30 CEST