Squid 3 memory usage
-
The code is absolutely the same… What's added there is to show you that IT WORKS. There's no new relevant code in there that'd fix the ghosts in your browser/pfSense box/god knows what.
-
Same error, (line 5 instead, but its the same code ;) )
The box is entirely stock apart from:
A few squid cache settings, Clam pointing at a regional update server, and your code for adding the widget, fixing the widget, adding the extra logs, and this.
Nothing else, it's a fresh install, and this has been duplicated on two boxes, both with fresh installs.As always, thank you for your input and suggestions.
-
For the moment don't worry, i'll spin up another test box, test on that to see if there's something odd.
-
Strangely, two new VM's, one the code worked on, the other it didnt.
The only difference between the two is post install they were given different external IPs.
Very odd, ah well. -
Well the watermark % variables appear to do nothing.
Swap is now at 100% on one of the boxes, with RAM at about 70% usage.
Top reporting Squid at 11Gb usage.This is with default settings in Squid too.
-
There are docs on the watermark stuff. When you look there, they basically say that it's sort of "best effort" configuration and that stuff will get ignored whenever Squid considers it necessary. Let me restate for the last time: upstream issues will not get sorted on pfSense.
And, heck… the docs are even linked from the GUI. ::) http://www.squid-cache.org/Doc/config/cache_swap_low/
Perhaps you are completely missing what "swap" means in Squid context. That's not the OS swap.
-
There are docs on the watermark stuff. When you look there, they basically say that it's sort of "best effort" configuration and that stuff will get ignored whenever Squid considers it necessary. Let me restate for the last time: upstream issues will not get sorted on pfSense.
And, heck… the docs are even linked from the GUI. ::) http://www.squid-cache.org/Doc/config/cache_swap_low/
Perhaps you are completely missing what "swap" means in Squid context. That's not the OS swap.
Cheers for the links, however im fully aware of this thank you.
To be clear, i am not asking you to fix upstream issues. I'm aware that there's little to no point with 2.3 on the horizon.
I'm simply reporting so others who may happen can see this is maybe not limited to them if they are having it.Once i've "tuned" it as well as i can i plan on listening what i've done that helps mitigate the issue so others can benefit in the interim.
-
Dude. Why are you referring to the watermark and your OS swap here, then? That stuff deals with disk cache object eviction. Not with evicting things from your physical RAM/swap.
Well the watermark % variables appear to do nothing.
Swap is now at 100% on one of the boxes, with RAM at about 70% usage.This has nothing to do with it. Simply sanitize your cache size if you are running out of memory. Or get more RAM. There are docs on the memory requirements as well.
-
Correct, but its the system swap, used by the kernel, that's getting used up.
Testing the last week or two has shown improvements to the memory and system swap usage based on tweaking these variables. "Trial and error". -
Yeah. So, either decrease your cache size, or increase your RAM size. Messing with watermarks is completely useless.
-
Cache size has already been experimented with, as has the RAM allocation.
No difference either way with memory usage.