On Thu, Apr 26, 2012 at 11:25 AM, Douglas Hubler <[email protected]> wrote: > On Wed, Apr 25, 2012 at 9:48 PM, Joegen Baclor <[email protected]> wrote: >> The issue I could think of would be cf-engine would detect the death of the >> proxy and would try to resurrect it. This can loop indefinitely until mongo >> is up. This si fine for as long as the resurrection loop would not eat up >> cpu and kill the system beyond recovery. I suggest that the first thing >> proxy init does is a mongo health check prior to creating media relay syms. >> This is the heavy part of proxy init. > > Sounds good, but you would need something to detect proxy is down and > send cfengine to watchdog mode, which...if you could detect proxy is > down in the first place, why not just start it back up. > > Also, this something we'll need to do for every service, so we need a > general solution like snmpd and monit. That's what these tools were > built to do, no need reinventing the wheel, we just need time to get > the working. > > To me, the watchdog is not the controversial part. The controversial > part is allowing the proxy to shoot itself in the head. Looks like > you are in agreement there.
I thought that proxy (mongo driver) is smart enough to elect another mongo instance running and to use it in order to complete the call (instead dying and call to be completed by other proxy). If no other mongo instance running then I agree proxy should die. George _______________________________________________ sipx-dev mailing list [email protected] List Archive: http://list.sipfoundry.org/archive/sipx-dev/
