by Elsewhere 36 Replies latest jw friends
that's all very well but if there's a glitch which takes up alot of the processor usage you will end up having reboots of the server due to high traffic counts. We have a product that is having similar issues on a HP-UX platform. Heartbeat monitors are all well and good but cause more grief than they solve
Steve, of the 'pissed off support engineer' class
install Debian Linux.
*hides smart ars smurk*
that's all very well but if there's a glitch which takes up alot of the processor usage you will end up having reboots of the server due to high traffic counts.
I think it would work. You wouldn't reboot/restart services the first time it failed. I'd do it this way... set a counter to zero. Every time the test is run, the counter would be incremented if the server failed to respond. The counter would reset to zero if the server responded correctly. If the counter exceeded a certain value, that would mean the server has failed to respond for a while. Then services could be restarted.
This may be a problem if the CPU is maxed out under normal circumstances and the server fails to respond at this time. In those normal cases you wouldn't want to do anything if the problem is going to remedy itself. I'd say that isn't usually a problem with this site because it always seems to respond pretty quickly unless the site is completely down.
Sens: I was hoping you'd see that... lol
I do have a process that monitors the web and will report if it's down (which is great when I'm asleep eh?!). It can be set to automatically reboot if required or restart the process - I'll have a look at configuring this.
I'll have a look at configuring this.
OpenService, ControlService, etc., will do what you need. Or you can just launch "net stop" and "net start"...
I prefer to use ADSI to recycle the app pool for the particular web
net stop / start would do it though.
