Re: [squid-users] squid not cachig big files.

Using FreeBSD 10.0-RELEASE amd64 and squid 3.3.11 via port
/usr/ports/www/squid33 on ZFS filesystem.

1392376616.882 SWAPOUT 00 000002A6 D27E641E51B22F49E25D124338E2D388
200 1392390889 1384782870 -1 video/mp4 245329905/245329905 GET
http://fs171.www.ex.ua/load/e291494a3e02be
e12326fbf54bee8c6e/82763797/Xonotic%200.7%20-%20Mossepo%20(POV)%20vs%20ZeRoQL%20-%20Silent%20Siege.mp4

1392376981.320 RELEASE -1 FFFFFFFF 487311FC22CB4BF63943CD89B8516174
200 1392391193 1391569953 -1 video/x-msvideo
843638784/843638784 GET http://fs75.www.ex.ua/load/571c72f6f
b348a1951a6a02d01c9205a/93458980/1x01%2602%20-%20Book%20of%20the%20Sword.avi

thanks for comments on the options refresh and cache_dir.

but as I understand it, it can not be a reason nostore items into cache?

15 февр. 2014, в 15:53, Amos Jeffries <squid3_at_treenet.co.nz> написал(а):

On 16/02/2014 1:31 a.m., Михаил Габеркорн wrote:

Good day!

my squid cant' cashing (store to cache) big files >200mb.

what i do wrong?

32-bit build of Squid?

32-bit operating system drivers? (unlikely, but maybe).

Any particular URLs being troublesome?

<snip>

cache_mem 1800 MB

maximum_object_size 999 GB

minimum_object_size 0

cache_dir ufs /var/cache 30000000 96 984 min-size=0 max-size=300000000

I'd make that L1 value 128 or 256 subdirectories instead of 96. And less
on the L2 subdirectories. The idea of L1/L2 is to flatten the lookup
time on the 2^27 objects Squid can put in there.

Also, note that Squid has that 2^27 object count limit. That is
absolute. It is no use having 30TB cache size if the objects inside it
only use 2TB.

Also note that Squid is best served by having cache_dir arranged one per
physical device. Drive partitioning on one spindle of spinning media
gives only problems, likewise RAID / LVM joining of devices for extra
large space leads mostly to problems.

<snip>

refresh_pattern . 900000000 80% 900000000 override-expire
override-lastmod reload-into-ims ignore-no-cache ignore-private
ignore-auth

Several seriously bad ideas in the above.

If you are using a Squid older than 3.4 please consider an upgrade over
doing that type of refresh_pattern. The latest Squid can cache a large %
more of traffic *properly* without:
a) leaking any users private information to the other users, or
b) damaging the way the pages display, or
c) "corrupting" web application processing.

They will also protect you from N-bit wrap on those time limits and from
exceeding the HTTP protocol storage BCP limits.

HTH
Amos
Received on Sat Feb 15 2014 - 14:23:49 MST