john at startupgiraffe
Jul 17, 2013, 9:55 PM
Post #1 of 1
(59 views)
Permalink
|
I am very curious about the oft quoted 5000 worker thread limit for varnish
instances. Particularly:
1) Is this recommendation still best practice despite having been made
years ago? Is it still relevant for -- now readily available -- beefier
server configurations like 16 or 32 cpu monsters with hundreds of GB RAM
and 10gb ethernet? Part of the original recommendation mentions bumping up
against file descriptor limits, but a cr1.8xlarge ec2 cluster instance
allows over 20m open file descriptors.
2) Is it the recommended limit for a single varnish instance, or a
recommended limit machine-wide. For instance, if I'm running 5 varnishd
instances on a single host, can each instance safely flex up to 5000 worker
threads concurrently or should the sum of worker threads across all
instances not exceed 5000?
Very interested to hear -- thanks!
Johnny
|
