Continuing Internet Issue

[thomaspark]

Hi Guys,

Here a good one for you all to get your heads around. We have just had installed a new network with 3 Virtual Hosts and HP ProCurve Switching throughout our network. We use a smoothwall for filtering and we are experiencing poor internet connection. We see our internet connection drop out several times a day for period of up to ten minutes but this does not affect all users. During these times we are able to ping bbc.co.uk and google.co.uk with the response time varying from 15ms to 96ms and sometimes we even see error pages generated by our proxy saying unable to forward request at this time. If we do a trace route all of our internal switches respond within 1ms including our proxy server.

On our network we use SCCM 2012 to deploy OS, Application and Software Updates. We also use a TMG Server to NAT through to our SWGFL assigned IP range, this was done because our IP range from SWGFL was not big enough. But again all of this kit responds in 1ms.

We have fibre to all of our cabs and 10 GB direct to our servers.

So my question is where is our problem?

[bossman]

We too have been experiencing slow internet speeds, done all the above to no avail.

 

I connected my laptop directly into our BT router using the external address of our smoothwall box and connected to the internet directly and what a difference, I could constantly ping BBC - Homepage and the TTL response times were constant at 16ms for a good while.

 

Next to check the latency of the line used command line to download a 50Mb file which took seconds, then used the Speedtest.net - The Global Broadband Speed Test site and this confirmed all was well with our BT line.

 

Next was to ping through the SW box with my network computer having created a exception and a port rule allow direct access, one would feel that this would be the same as the direct connection earlier but no! as soon as other connections were made through the SW box it began to drop TTL response times drastically from 16ms up to 350ms at one point and this was with the postgres database stopped. As soon as I start the database it is even worse and the WA goes up to around 20 and as I have already said when the database is stopped the WA still rises to 3 and 4.

 

Our intranet site can sometimes take up to 5 seconds to load although I have reconfigured our authentication back to its original staus of default ldap as opposed to NTLM and this seems to have made a slight difference.

 

All in all i feel it is our SW box and i hope to have a conversation with the guys tomorrow at the edugeek conference, I hope you reading this Tom.

 

Anyway hope to have it all sorted soon.

[tom_newton]

@robk is sitting next to me and pointed me at this. If you have a lot of wait states, it may be the cache performance I'd suffering. Suggest a call to support (will call for you tomorrow if you catch me?) if the box is under serious disk load I suppose it could affect ping times too. Needs proper investigation mind.

[thomaspark]

Hi Guys,

 

We go through SWGFL for our Internet connection and keep seeing realy slow internet and slow DNS Lookup, with sometimes the internet cutting out. Sometimes we are lucky to get an error generated by the upstream proxy which changes.

 

Kind Regards

[MikeGilbert]

Hi Guys,

 

We go through SWGFL for our Internet connection and keep seeing realy slow internet and slow DNS Lookup, with sometimes the internet cutting out. Sometimes we are lucky to get an error generated by the upstream proxy which changes.

 

Kind Regards

 

I need to be careful how I phrase this... I know a potentially similar issue caused, as far as can be determined, by the level of bandwidth purchased by an LA being exceeded by the aggregated demand of the schools within the LA. It may be worth investigating along those lines, as it's a situation that could be repeated elsewhere if the demand forecast is being exceeded, which it might well be.

[bwestlake]

swgfl suck...

[WinNet]

Hi,

SWGFL have been investigating our internet connection issues and have not found anything wrong they have said that all of their kit is working perfectly it must be something wrong with out network or a change that we have made. Well we haven't changed anything and our network responds on a Trace Route in under 1ms where as once it's on the internet it takes upwards of 16ms to 110ms. Any idea's on how would be best to prove our network is working correctly?

 

Kind Regards

[MikeGilbert]

Hi,

SWGFL have been investigating our internet connection issues and have not found anything wrong they have said that all of their kit is working perfectly it must be something wrong with out network or a change that we have made. Well we haven't changed anything and our network responds on a Trace Route in under 1ms where as once it's on the internet it takes upwards of 16ms to 110ms. Any idea's on how would be best to prove our network is working correctly?

 

Kind Regards

 

Once upon a time there was a thing called Cricket that gave a full log of schools' connectivity to SWGfL, to which I managed to acquire a login. It's not there anymore, sadly, but there is a replacement system. To which I don't have access. It may be worthwhile approaching the problem from a slightly different angle, and asking if you can have access to your usage data to enable to optimise the way you're utilising your bandwith.

[stevetheb]

http://service.swgfl.org.uk/cacti

[MikeGilbert]

http://service.swgfl.org.uk/cacti

 

[h=1]Not Found[/h]The requested URL /cacti was not found on this server.

---

Apache/2.2.3 (Red Hat) Server at service.swgfl.org.uk Port 80

 

[stevetheb]

Access it from here - with a login...

 

SWGfL Electronic Service Interface

[SchoolsBroadband]

I'd also recommend doing a pathping command.

 

This essentially does a traceroute (tracert) but also sends loads of ping packets to each hop along the way to the destination. You should be able to tell from here at what point the issue is happening along the link and if its the LA or something local. it'll also show you if there are any lost packets.

 

Dave

[Geoff]

I'd also recommend doing a pathping command.

 

It would be best to do the diagnostics from the smoothwall box surely. @tom_newton Do the smoothwall boxes come with mtr installed for example?

[tom_newton]

No they don't - will have a look at packaging it as long as it has no dubious dependencies.

[Geoff]

No they don't - will have a look at packaging it as long as it has no dubious dependencies.

 

root@woody:~# ldd /usr/bin/mtr
linux-vdso.so.1 =>  (0x00007fff07d79000)
libresolv.so.2 => /lib/x86_64-linux-gnu/libresolv.so.2 (0x00007fe5cd420000)
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fe5cd124000)
libncurses.so.5 => /lib/x86_64-linux-gnu/libncurses.so.5 (0x00007fe5ccf02000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fe5ccb43000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fe5cc93f000)
libtinfo.so.5 => /lib/x86_64-linux-gnu/libtinfo.so.5 (0x00007fe5cc717000)
/lib64/ld-linux-x86-64.so.2 (0x00007fe5cd644000)