Willy, I can see why, with some web farms, you wouldn't want to take servers out of rotation after just one, or a few 5xx errors, particularly since often they are caused by bad user input. In our case, any 5xx errors are almost always an indication that the server in question is in a bad state. Particularly problematic is that a server serving 5xx errors tends to do so much faster than one responding to legitimate requests. This means that a bad server can serve several thousand requests before the next health check kicks it out of service.
Implementing inline monitoring dropped our 5xx error rate by two orders of magnitude, so it is pretty important for us. If we move forward, we'll likely submit a patch if the functionality doesn't exist as things stand now. Perhaps it would be better if it was a counter that took a server out after a set number of consecutive failed requests. Jonah On 5/24/09 9:56 PM, "Willy Tarreau" <[email protected]> wrote: > Hi, > > On Fri, May 22, 2009 at 11:37:14AM -0700, Jonah Horowitz wrote: >> I¹m currently testing HAProxy for deployment. Right now we use NetScaler >> load balancers, and the provide a feature called ³inline monitoring². With >> inline monitoring the Netscaler will take a server out of rotation if it >> responds with a 5xx error to a client response. It does this separate from >> standard health checks. Is there a way to do this with HAProxy? > > No, and I don't want to do the same as it seems a little bit risky to me. > However what is planned is to switch to fast health-checks when a number > of 5xx errors is encountered. That way, it would significantly reduce the > time to detect a server failure without the risk of taking a server out of > the farm on random errors. > > Regards, > Willy > -- Jonah Horowitz · Monitoring Manager · [email protected] W: 415-348-7694 · F: 415-348-7033 · M: 415-513-7202 LookSmart - Premium and Performance Advertising Solutions 625 Second Street, San Francisco, CA 94107
smime.p7s
Description: S/MIME cryptographic signature
