What happens when perlbal reverse proxy backend server dies?

Can Sar csar at stanford.edu
Thu Aug 30 20:16:28 UTC 2007


Hi,

I have perlbal set up as a reverse proxy to load balance between two  
instances of Apache. When one of the machines is taken offline/dies/ 
etc. it seems that Perlbal becomes far slower at answering even a  
single request (no other load on the slower), and that load time goes  
from 1 second to 6 seconds or so. It seems to me that it tries to  
connect to the machine, eventually times out and then goes to the  
other machine but never marks the machine as down or anything else.  
Is this the expected behavior?

I would ideally like it to detect that the machine is down, not give  
requests to it and then eventually probe if the machine is back up  
after a few minutes or so. I guess having to manually re-add the  
machine after failure would also be ok.

I've tried to find documentation on the managment interface but  
haven't had any luck and the help command only lists commands without  
explaining their syntax...

Here's my config file:

CREATE POOL apache
   SET nodefile = /etc/perlbal/nodelist.dat

My CREATE SERVICE dynamic
   SET listen = 0.0.0.0:80
   SET role = reverse_proxy
   SET pool = apache
   SET persist_backend = on
   SET persist_client = on
ENABLE dynamic

Thanks!
Can Sar


More information about the perlbal mailing list