search

inevity / lusca-cache

Automatically exported from code.google.com/p/lusca-cache
0 stars 0 forks source link

Range requests HIT if file exist in cache, but later on - no #55

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
What steps will reproduce the problem?
1. Probably corrupted cache. On fresh cache it works fine.
2. Windows update clients which uses Range requests 
3. Download by wget file

What is the expected output? What do you see instead?
I expect that range requests will get also HIT. But they won't.

What version of the product are you using? On what operating system?
Linux, LUSCA HEAD, latest from SVN

Please provide any additional information below.
I download file, and i can see it is in cache
1251764677.615 112159 194.146.153.18 TCP_HIT/200 36225299 GET
http://msgr.dlservice.microsoft.com/download/0/3/F/03FA4580-3ED7-4D60-8CAD-17B99
88F0E39/en/wlsetup-cvr.exe
- DIRECT/208.111.128.59 application/octet-stream

At same time
1251764789.669    791 172.16.101.98 TCP_MISS/206 9242 GET
http://msgr.dlservice.microsoft.com/download/0/3/F/03FA4580-3ED7-4D60-8CAD-17B99
88F0E39/en/wlsetup-cvr.exe
- DIRECT/208.111.128.58 application/octet-stream
1251764789.901   3162 172.16.101.29 TCP_MISS/206 18322 GET
http://msgr.dlservice.microsoft.com/download/0/3/F/03FA4580-3ED7-4D60-8CAD-17B99
88F0E39/en/wlsetup-cvr.exe
- DIRECT/208.111.128.59 application/octet-stream
1251764789.993    567 194.146.153.62 TCP_MISS/206 7098 GET
http://msgr.dlservice.microsoft.com/download/0/3/F/03FA4580-3ED7-4D60-8CAD-17B99
88F0E39/en/wlsetup-cvr.exe
- DIRECT/208.111.128.59 application/octet-stream
1251764790.078   3745 172.16.152.242 TCP_MISS/206 10078 GET
http://msgr.dlservice.microsoft.com/download/0/3/F/03FA4580-3ED7-4D60-8CAD-17B99
88F0E39/en/wlsetup-cvr.exe
- DIRECT/208.111.128.59 application/octet-stream

If cache is fresh - it will be HIT. If it is crashed few times, it will get
HIT only on complete retrieval, but will get MISS on /206.

I can't attach cache.log with debug_options ALL,3 here, since it is large,
i will put in on my hosting as
http://www.nuclearcat.com/lusca-cache.log.bz2 (uncompressed 1.6GB)

http://www.nuclearcat.com/squid.conf.lusca is my config

Original issue reported on code.google.com by nuclear...@gmail.com on 1 Sep 2009 at 12:50

GoogleCodeExporter commented 9 years ago 
Let me know if you need any additional info. I can provide remote access to 
system,
but it is semi-embedded and 128Kbps only for ssh (thats only i have over low 
latency
link).
I can run with any other debug options...
I am very curious what it can be...

Original comment by nuclear...@gmail.com on 1 Sep 2009 at 9:41

GoogleCodeExporter commented 9 years ago 
Probably this issue linked with collapsed_forwarding on and
download_fastest_client_speed on

After disabling this two options my cache ratio improved.

Original comment by nuclear...@gmail.com on 3 Sep 2009 at 10:48

GoogleCodeExporter commented 9 years ago 
ok. Well, range requests aren't cachable. Full requests are; range requests 
aren't. Lusca (and Squid) will return 
a range reply if the full object has been cached.

download_fastest_client_speed doesn't yet do anything. :)

collapsed_forwarding being 'on" has resulted in various large objects becoming 
uncachable in the past. I need 
to actually sit down and document how that stuff works, including the "weird 
crap" that I've seen when I've 
used it in deployment. I'm tempted to make collapsed forwarding ACL driven, but 
the one thing Lusca doesn't 
need right now is even more ACL driven options slowing down processing.

Original comment by adrian.c...@gmail.com on 12 Sep 2009 at 12:51