On Fri, Feb 25, 2005 at 02:35:44PM +0800, Bosco Lau wrote:
I have been using Apache 1.3.xx under Ultra Monkey HA/Load balancing
package (load balancing thru NAT). Everything seems to work perfectly
fine... Until I upgraded to Apache 2.0.53 (prefork). Seems that the
'pinging' by the ldirectord is causing problems to the Apache2 server.
But the even stranger thing is that when I connect directly to the
Apache2 server without going through the load balancer everything seems
to be fine (I have even tested it with some stress benchmarking).
I have tested the following conditions:
Apache 1.3 without LB (load balancing) - OK
Apache 1.3 with LB - OK
Apache 2.0.53 without LB - OK !!!
Apache 2.0.53 with LB - absolutely NUTS !!!
I suspect that Apache2 is reacting strangely to the negotiating/request
by the ldirectord, which requests and reads a test file from the apache2
document root and try to compare the content of the test file to that in
the config file of the ldirectord. But there isn't such a problem with
Apache 1.3. At the moment the ldirectord is configured to make a request
to the apache every 1 sec (which is not a high loading at all). My hunch
is something to the with the Apache2 configuration of the
multi-threading (???). Maybe the threads from the 'pings' are killing
the server?
Please advise in anything I can do to the Apache2 or the ldirectord
configuration, or anything else for that matter.
That sounds really weird. Could you try using a connect check
rather than a negotiate check and see if that makes any difference.
