Sudden spikes cause varnish to crash

[chrism2671]

I'm not 100% sure happened here. This is running Mercury 1.0 on our production site. It seems at 1605 there was a sudden spike in load. I am not certain from the logs exactly what caused the spike. The result was that varnish died, and consequently the site went offline (see cloudwatch screenshots). FYI we are using RDS for the database.

I would like to suggest employing some self healing in the build; something like monit that watches for unresponsive processes and can restart them, or in the worst case, initiate a reboot.

What do you think?

[chrism2671 (chrismuktar)]

I've just tried this on a 32-bit small instance AWS AMI 1.0 and found something similar, even with no load. It was fine, when I came back a few hours later, varnish had died, without log. I'm not entirely sure what's causing it (I know that early versions of varnish had stability issues). As this is a stock, unmodified instance, I imagine that others might see the same. If anyone has any suggestions, I'd be very appreciative.

Thanks,
Chris.