Caddy uses lots of ram

Caddy 0.9.4 seems to use way more memory than 0.9.3 did:

This is with low/moderate load, 4 domains & max. 3 subdomains per domain. Seems kinda high to me if you compare it to other servers

Update: RAM usage just doubled.

(The CPU usage is just unfortunate timing. There were multiple accesses at that time & it went back down immediately, ram usage still the same)

What are you using Caddy for? What’s your Caddyfile and what kind of requests are coming in, or what kind of clients?

Right now, Caddy serves a few static pages & acts as a reverse-proxy for a few services.
Most traffic is going to the services right now, one being an instance of Nextcloud (Owncloud) . it’s just a proxy statement with the transparent preset.

Caddyfile for the static pages looks like this:

root /srv/www/siteroot
fastcgi / /run/php-fpm/php-fpm.sock php
ext .php .html

And for the proxies:

root /var/tmp
proxy / my.local.ip transparent

The memory usage went back down to a couple hundred MBs, i suspect the amount of traffic going over the proxies caused it to spike - this may be worth investigating further

@coderobe Can you file an issue on GitHub? That way we won’t forget to have someone investigate. Please include enough information to reproduce the high RAM usage in a test environment (meaning full Caddyfile, not just little snippets, and include instructions to follow to see the RAM use go up in a definite way, not just “something like this” if possible).

1 Like

Slightly off topic,

That works? I’ve always used

proxy / my.local.ip {

I was assuming @coderobe was just paraphrasing but if that is literally in the Caddyfile, then that could be a problem: Caddy would treat that as an upstream address.

1 Like

I’m actually using

i was just shortening the response, sorry for the confusion :slight_smile:

1 Like

I’ve posted an issue on github as requested:


Thanks Robin! Someone will get to it :slight_smile:

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.