Search squid archive

Re: AW: Client randomly reloads pages every second - cache/parent issue?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 27/11/2012 8:53 p.m., Stefan Bauer wrote:
-----Ursprüngliche Nachricht-----
Von:	Amos Jeffries
On 6/11/2012 10:39 p.m., Stefan Bauer wrote:
Dear Users/Developers,

we're having the following squid structure in our environment:


Parent-Proxy1+2
|
DMZ
|
Child-Proxy1+2
|
|
Clients


We notice quite often the case where the client/browser is automatically
starting to refresh a random page every second for unknown reasons.
We're unsure if this is triggered by a wrong configuration or by a bug in the
client itself.
How can we debug this further?
Squid has no control over the client/browser choice of when to
re-request an object. The URL you presented results in a *private*
response so both proxies are supposed to have nothing to do with it
beyond network fetch and delivery.

The URL also has several other major problems which could be screwing up
the browser:
http://redbot.org/?uri=http://www.mcseboard.de/windows-7-forum-76/kein-speichern
-credentials-moeglich-win7-client-term-server2008-169219.html


This is from the squid cache log:

1352194376.779    224 192.168.0.5 TCP_MISS/200 14432 GET
http://www.mcseboard.de/windows-7-forum-76/kein-speichern-credentials-moeglich-w
in7-client-term-server2008-169219.html miw
ROUNDROBIN_PARENT/parentproxy1.intern.domain text/html
1352194377.013    214 192.168.0.5 TCP_MISS/200 13900 GET
http://www.mcseboard.de/windows-7-forum-76/kein-speichern-credentials-moeglich-w
in7-client-term-server2008-169219.html miw
ROUNDROBIN_PARENT/parentproxy1.intern.domain text/html
1352194377.380    188 192.168.0.5 TCP_MISS/200 14173 GET
http://www.mcseboard.de/windows-7-forum-76/kein-speichern-credentials-moeglich-w
in7-client-term-server2008-169219.html miw
ROUNDROBIN_PARENT/parentproxy2.intern.domain text/html
1352194377.622    212 192.168.0.5 TCP_MISS/200 13900 GET
http://www.mcseboard.de/windows-7-forum-76/kein-speichern-credentials-moeglich-w
in7-client-term-server2008-169219.html miw
ROUNDROBIN_PARENT/parentproxy2.intern.domain text/html
We setup this in the child-proxies config:

cache_peer parentproxy1.intern.domain parent 3128 0 round-robin no-query
cache_peer parentproxy2.intern.domain parent 3128 0 round-robin no-query

Could this be the problem?
I don't see any evidence of a problem in the proxies. It all appears to
be browser and server issues.
Hi Amos & the mailing list,

sorry for my late response. thank you for your answer. Are you sure, it could not be caused by the round-robin option to the cache_peer parents?

*could* be yes. *if* the website were relying on clients always having one IP through their visit. As demonstrated by its very broken uses of Vary, Pragma, Range, compression and 304 in responses it could very well _also_ be broken in its handling if HTTP statelessness.

Could also be something in the pages themselves. There are half-broken little hacks like "top.location.href = location.href;" in a few places that play with page loading locations.


I have reports from several departments that this was not happening with the single instance proxy we had previously.
now we have this random reloads of pages - causing for example several hundred thousand hits on some machines :/

List all the differences; single-instance is only one variable, version is another, feature support changes, config changes etc are all potentially relevant.

Amos


[Index of Archives]     [Linux Audio Users]     [Samba]     [Big List of Linux Books]     [Linux USB]     [Yosemite News]

  Powered by Linux