Hi guys!

This is a very heavy prob I'm having at school (and sharing with 1600 people )
I don't know if this is a Linux problem or whatelse, but we run Linux servers, so I'll post here. Sorry if I'm wrong!
I'm an admin at our school. We're running Debian servers, with well- configured squid and all that stuff (proxy for http->port 8080).

We have switches in every room for the network. There were NO problems. For some reasons we switched to layer3 switches, 8080 is open both tcp and udp for browsing. 21, ssh and some others are open too (need them for some apps...), but we have problems with browsing.
It takes ~2 minutes to display a page. Internet isn't really slow, downloading is very fast. But the browsers let you wait for a few minutes after loading the page. Sometimes the webpages load a little bit, then stop, then load and so on...
/etc/squid.conf is configured to use 8080 for http!
The internal network (for example the homepage on one of our servers) loads extremly fast. No prob here.

Well, on our other servers "directly" connected to the gateway internet is fast as usually (our other servers).

Any ideas?
Thanks for reading this!

-Marvelloard

Are your DNS servers internal or external, and what network connections exist between your Squid box + the DNS servers. This is an option within squid on reverse DNS lookups I think that can slow things down if enabled. Also, are there any other networking issues - I've found firmward updates on switches can resolve matters, but this seems more Squid related. From a client, try doing an nslookup to see how quick the name is resolved, and then try pinging by domain name and see how quick is resolves. Further (quite a few things here!), is the problem across all browsers - have you tried alternative browers, including remove the proxy server from the connections settings and connecting direct to the Internet and seeing how quick page load?

That should you going for a while

Hi!

Thanks for that great post, I'm currently at home, but I will post more details at school tomorrow!
Yes, I tried different browsers on different OS'es (SuSE, Windows, Firefox, Opera, Netscape, IE), it's all the same.
And yes, sounds like a squid problem, but as I said, before layer3 switches everything was working well. And is still working well in the rooms still equipped with our old layer2 switches (sorry I forgot to write down that!).

Thanks again, I'll write back tomorrow at school!

Okay, let us know what happens. If the other machines running off the older layer 2 switches are fine, possibly move away from the idea of it being a Squid issue.

Also, you don't have anything extra enabled on the new layer 3 switches such as STP or link aggregation do you? I can't imagine moving between layer 2 > layer 3 switches causing problems other than additional functions on the newer switches.

Hi

I didn't get the chance to try everything today. Sorry but I'm a bit busy, I'm doing this admin job in my free time and I have a lot of other things to do (projects etc..).
But here's something:

Layer3 switches are running the latest firmware, and there's nothing special enabled. It can't be a squid prob I think, and I don't think it's a switch problem. Why? Please read on.

I've made a little painting of our network.
http://www.htlwrn.ac.at/~wb/graphics/htlwrn.PNG
As you can see, there are two Networks: The Qbe network, and the EDVO nw.

Some explanations:
1)Computers in Qbe network behind L2/L3 switches are able to access XTC server very fast (internal, blue line)
2)Computers in Qbe nw behind L2/L3 switches are able to access edvo-srv server very fast (http, internal, orange line)
3)Computers in EDVO nw behind L2/L3 switches are able to access the internet very fast (using our INTERNAL DNS, gray line)
and finally
4)Computers in Qbe nw behind L3 switches have problems accessing the internet using INTERNAL DNS (red line)!!

But please keep in mind:
-with L2 switches internet works well in Qbe nw
-the L3 switches are EXACTLY the same in both nws (hardware, firmwares, even configuration files!!!!!)

We're gonna exchange the two cisco ports for the two networks soon, one of us had the idea that the cisco router could be buggy (whatelse could be the problem...?).
But this could take some time, because we don't have access to the other server room permanently, and the admin there is sometimes _very_ unfriendly (maybe good chances th get access, christmas is coming...

Any ideas/thoughts/arguments/etc?
Thanks!
-Marvelloard

Looking at your diagram, the Cisco router shouldn't have any effect on DNS queries from the QBE network as you have a direct connection from your L2/L3 switches to the DNS server. The only thing that springs to mind is how your network addressing is setup and is your DNS configured primarily for one network segment, or are you running dual network cards in the DNS server to handle queries from the two incoming network connections? If so, could it be the NIC handling this segment is at fault? There does seem to be quite a cross of cables and systems there and I don't quite understand why they're linked quite the way they are.

But, should this not all be the job of the network admin that's in charge of it anyways, or am I missing something ?

Sorry for the late reply, but I had/have a lot of work to do.
Things changed a little bit: Students in the EDVO network are having the same troubles. They just were too stupid to tell anyone ARGH!
There is an update for the router available, we'll upgrade soon.

Yes, it should be his job (I'm just the php guy , but he doesn't work in a L3 room, so he doesn't have any problems->he doesn't care...

Anyway, thanks for the help, and Merry Christmas!
I'll be back =)