It’s two HP workstations (Eric is an xw6400, and Ernie is an xw6200). Originally I intended buying just the xw6400, but I cheaped out and went for the xw6200 instead, thinking it would happily run ESXi 5. I was right, it does run it. The trouble is the processors don’t have the Intel vt-x extensions, so they can’t run 64-bit virtual machines. I mulled over my options, sold some stuff on eBay, and decided to get the xw6400 as well, as it DOES support vt-x. It actually worked out quite nicely as it now means I have 16 gig of RAM between two machines, a reasonable amount of disk, and the ability to spread loads. Basically the plan being to run 32-bit stuff and clients on Ernie, and anything that needs 64-bit can go on Eric.

You’ll notice two vSwitches on the boxes. What I have is two Broadcom NICs in each box. One connects to my local LAN, the other two connect back to back with the Virtual Machines vSwitch on them. All the VMs connect to this switch, save for one, which runs m0n0wall and bridges onto the main network too. It’s configured just to route, and do not much else; my main router (running DD-WRT) has a static route pointing to it, so clients on the Virtual Machines vSwitch are able to get out onto the internet, but not send any wayward data onto my main LAN.  I may change m0n0wall it for something a little more suitable in time, but for now it works. I’d like something which is a routing firewall, rather than a slightly knocked around NAT gateway.

Finally, there’s an NFS filesystem exported from my NAS with ISO images on from Technet and elsewhere, so I can install software and not waste space on the local datastores.

So far I’ve only thrown a few VMs on, but it seems to work quite nicely, and they aren’t particularly noisy when running. I only turn them on as needed, so power isn’t a big concern either. Hopefully with this setup I can play around with stuff to my heart’s content.

If you’re wondering where I picked up this stuff, it was from a company on eBay called bargain*hardware selling ex-corporate workstations. Pretty good prices for the kit you can get.

Eric and Ernie (and the NAS)

My problem was I’d spend all day at work maintaining boxes, then come home and occasionally have to maintain another box. Not only that but I was paying the best part of £10 a month for this. It just suddenly didn’t add up any more. I had a look for some new hosting and was recommended iPage.

Their support was excellent, and their package was very generous (for the price). The problem was, their hosting is in the US in Boston. It seems their routing and peering is really not up to capacity, so page load times were horrendous. They actually use the caching system Akmai to give better page load times, and it’s a freebie for UK customers; in hindsight, that should’ve set off the alarm bells. With dynamic content like a WordPress blog, it’s really a pointless endeavour using caching as it was the initial connection and page load which took time, not pulling down static content like images. Although the images were slow too they weren’t as noticeable as staring at a blank page for anything up to ten seconds. I also host the bnluk.net site, and it just wasn’t good enough, as it was frustrating users.

The final nail in the coffin came earlier this week when their database server decided to hose itself, and my databases with it. It came back up, seemingly OK, but the databases weren’t working properly. I resurrected it and then started looking for new hosts.

After a great many recommendations from various people I found myself back at UKWSD, signing up to a slightly higher tier of service than I’d had before, which gave me the extra scope to host what I wanted. The account was set up in no time, all the data slurped over, and the DNS was switched. iPage refunded me my hosting charges.

If I was hosting a site aimed at US-based users, then I’d possibly use iPage again, but for UK-based users it’s a no-go. It’s a shame because as I say, their support and general service was great. Just a shame their network isn’t.

Rapidswitch were in the throes of migrating all their VPS customers to a sister company’s platform, and my cancellation request seemed to have dropped through the cracks, so I now need to unravel that too. I still need to tidy up my domain registrations – they’re between two hosts still!

But, I DO have two working domains. So that’s the main thing. All thanks to UK Websolutions Direct.

If you’re installing the Python Cheetah library, you’ll most likely follow the instructions, which say to do the following:

To install Cheetah in your system-wide Python library:

1. Login as a user with privileges to install system-wide Python packages. On POSIX systems (AIX, Solaris, Linux, IRIX, etc.), the command is normally ‘su root’. On non-POSIX systems such as Windows NT, login as an administrator.
2. Run python setup.py install at the command prompt.
3. The setup program will install the wrapper script cheetah to wherever it usually puts Python binaries ("/usr/bin/", "bin/" in the Python install directory, etc.)

Quite simple. However, what it neglects to mention is that there is a little module within the library called ‘namemapper’. What it does, I don’t entirely know (something to do with mapping names would be my guess), but this module is written in C. Simply running the install as per the Cheetah docs won’t build this module, which results in much slower performance from Cheetah overall as it falls back to a python-based internal equivalent. However, if you run the command python setup.py build BEFORE the install command, it will happily compile said module. The install will then install it along with everything else, and all is well. Obviously you need gcc installed to do this (I had the dev pattern installed on my OpenSuSE box).

Why the docs don’t mention this, I don’t know. Still, nice to know there’s a simple fix.

Because of the way it works, it’s possible to take any input, and spit it out in any way pretty much. Some time ago I’d seen a discussion on a mailing list about using VLC and a digital TV card to broadcast live TV over the network (using multicast). I’d had a quick play, but never got it working. A discussion popped up on the Multiplay forums, so I thought I’d try and get it working again. Last night I was successful, so I thought I’d blog it up in case it’s useful for others. VLC is documented, but the documentation is a bit hit and miss in places – always the problem for many Open Source projects.

Read more after the jump.

Background

Digital TV, using either satellite (DVB-S) or terrestrial (DVB-T) is fundamentally the same – MPEG2 Transport Streams, which are multiplexed into one stream, then broadcast on a single frequency. The tuner will tune to the frequency, then pick the stream it wants, and play it on screen.

The idea behind what we’re doing here is to take that multiplex, and pick out the streams we want, then just multicast them as is onto the network. Using the SAP protocol, it’s possible to announce them in a nice neat way so clients can pick them up easily.

Because it’s already MPEG2, no re-encoding is really needed. I suppose you could do that, but I’ve not tried. This method uses hardly any CPU time, as it’s just pulling data in, patting it about a bit, and throwing it out the other side. In the original mailing list postings I read, the author was using some old Pentium 4 PCs with hardly any RAM. They were capable of tuning multiple channels with ease.

One caveat of this is that the tuner can only lock onto one frequency, so one multiplex is all you can play with; on Terrestrial (i.e. Freeview), the BBC multiplex contains all the BBC channels. If you wanted to multicast BBC and ITV channels for example, you’d need to add a second tuner and tune to the ITV multiplex.

In my investigations, I had the following setup:

Both are connected over a Gigabit network. However, the video streams are quite small (5-6Mb/s). Even 10Mb network could cope with that, although not many concurrently!

The server end

First of all, here’s the full command to kick start VLC on the Server.

vlc dvb-t://frequency=578000000 :adapter=0:dvb-bandwidth=8 --programs=4171,4235
--sout #duplicate{dst=std{access=udp,mux=ts,dst=239.255.255.239,sap,name="BBC One",
group="BBC"},select="program=4171",dst=std{access=udp,mux=ts,dst=239.255.255.238,
sap,name="BBC Two",group="BBC"},select="program=4235"

That’s all on one line, I had to break it up a bit!

So an explanation of it. Basically it runs VLC, tunes the frequency, using a selected adapter and channel bandwidth. It then selects the two programs I want (BBC One and Two), then outputs them via UDP on two different multicast addresses, using SAP to announce it.

First of all, I’d recommend running VLC up manually. Go to Media >Open Capture Device. Change the capture mode to DVB Directshow. Select the video format you’re using (I’m using DVB-T). Then dial in the frequency of the multiplex you want. Select 8Mhz for the bandwidth. You can find out the multiplex frequency using UKfree.tv.

Click Play, and it should open.

If it doesn’t you might need to select your video adapter explicitly. I have two in my PC. The DVB-T card is adapter 0, whilst my DVB-S card is adapter 1. Click Media > Open Capture Device again , then tick Show more options and enter “:adapter=0” into the options box. In the previous screenshot I have highlighted this. Change the number until it works.

Once you are able to see the video, go to Tools > Messages. Change the Verbosity to 2. Stop and start VLC and watch the log fill up. There’s too much to try and work with it in the window, so click Save as…. Open the file you have saved, and look for a section similar to the following:

ts debug: new SDT ts_id=4107 version=3 current_next=1 network_id=9018
ts debug:   * service id=4171 eit schedule=1 present=1 running=4 free_ca=0
ts debug:     - type=1 provider=BBC name=BBC ONE
main debug: EsOutProgramMeta: number=4171
ts debug:   * service id=4235 eit schedule=1 present=1 running=4 free_ca=0
ts debug:     - type=1 provider=BBC name=BBC TWO
main debug: EsOutProgramMeta: number=4235
ts debug:   * service id=4415 eit schedule=1 present=1 running=4 free_ca=0
ts debug:     - type=1 provider=BBC name=BBC NEWS
main debug: EsOutProgramMeta: number=4415
ts debug:   * service id=4479 eit schedule=0 present=1 running=4 free_ca=0
ts debug:     - type=1 provider=BBC name=BBC Red Button
main debug: EsOutProgramMeta: number=4479
ts debug:   * service id=4671 eit schedule=1 present=1 running=4 free_ca=0
ts debug:     - type=1 provider=BBC name=CBBC Channel
main debug: EsOutProgramMeta: number=4671
ts debug:   * service id=4351 eit schedule=1 present=1 running=4 free_ca=0
ts debug:     - type=1 provider=BBC name=BBC THREE
main debug: EsOutProgramMeta: number=4351

Notice all the Service IDs. As you can probably tell, each of these represents a TV channel within the multiplex. You need to note down the ones you want to multicast. In this case I want BBC One (4171) and BBC Two (4235).

The next part of the command selects those programs and outputs them:

--programs=4171,4235 --sout

The next part is to make up the multicast. As the video is already in MPEG2 we don’t need to go re-encoding it, so it’s simply a case of spitting it out.

I’ll be honest, I’m not sure what this next section does – I think it is basically saying take the input and duplicate it to output elsewhere (i.e. other than on screen).

#duplicate{

We are going to broadcast on a multicast address (anything in the range 224.0.0.0 to 239.255.255.255). You need to choose your addresses. On a large network with multicast enabled, these could be in use, so it’s worth checking first. If you know you don’t have any multicast I think it’s fairly safe to pick a random address. I chose 239.255.255.255.239 and went down from there!

This next line is where we select UDP, make it a Transport Stream, select the target IP, and advertise it with SAP. The channel name, and group name are set here too:

dst=std{access=udp,mux=ts,dst=239.255.255.239,sap, \
name="BBC One",group="BBC"},select="program=4171",

Repeat for BBC Two. Use a different multicast address though!

dst=std{access=udp,mux=ts,dst=239.255.255.238,sap, \
name="BBC Two",group="BBC"},select="program=4235"

Now we have a full command, it’s a case of running it. I used a command prompt and ran VLC from there, but you could of course put it into a shortcut, batch file, registry run key, or whatever.

Once it’s running, a very simple, hacky way of checking it is working is to look at your network switch. If it’s flashing away merrily on all ports, then it’s doing something!

The clients

Run VLC on another PC, Go into Preferences > Playlist > Services Discovery and tick SAP. You might need to re-start VLC. Go to the playlists, and you should see a group with the name you made, and your channels. Click one, it plays!

As you can see, I’m doing this on my Mac, so it’s truly cross platform!

I did find in testing that it caused an effective DoS on my Wireless. I’m guessing the multicast was too fast for the slowest wireless device associated to the AP, and just wiped the lot out or something. Wired carried on working fine though. Something to bear in mind anyway!

I’m sure it’s possible to refine this a little more. The original version of this I saw had VLC daemonizing itself, which I’m not entirely sure is possible in Windows. Still, it’s a great way to get video served up over the network.

Oh, and remember, you still need a TV licence!

Via some spare fibres and various spare switches, we’ve got access to it in various switch closets. I wanted to be able to access it from my desktop. My preference was to use a virtual machine, which I could set up with all the software I needed and test. This would mean bridging said virtual machine to that network.

I added a second network adapter to my PC, unbound it from all the local protocols (so my PC doesn’t touch that network) and set VMware Player to use bridged mode. With VMware workstation, you can use the virtual network tool to configure which NIC bridges to the virtual machines. However, that tool is missing with VMware player. By default, it was bridging to my first NIC, which was no good, as it was on the uni network.

A little bit of clicking around, and I found that there is a protocol bound to all the network cards called VMware Bridge Protocol. By default, it seems to bind itself to all the NICs. I simply unbound it from my main NIC:

And left it bound to the NIC connected to the ADSL network:

I restarted VMware Player, and my virtual machine nows bridges to the ADSL connection.

Post update, it decided to redirect me to google.nl all the time. No idea why, as it had been working fine previously. Checking every option I could within Opera, I could find no reason why.

The answer, lies in a config file, which lives here: (I’m on Windows 7, so yours may vary slightly)

C:\Users\<Your Username>\AppData\Roaming\Opera\Opera\operaprefs.ini

Towards the end of the UserPrefs section, is this line:

Google TLD Default=.google.nl

That was the one! I changed it to .co.uk, restarted Opera, et voila, it now goes where it’s meant to. I can only assume it’s something the installer did. My system is set to UK everything (I checked) and I’ve never had anything else go whizzing off to google.nl. Even going to google.com redirected properly.

Very odd!

The thing about the call which intrigued me the most was that BT are notoriously bad at acting on changes of information. Indeed, somebody from BT told me that a lot of their billing hangs off company names and addresses, so changing it causes all sorts of headaches for the billing department – hence why they don’t!

I can’t help but feel my friend on the phone was cold calling from some random agency, basically looking to find info about us and what we do and needed, which they’d then cold call us about and try to sell to us. Maybe I was being over-cautious, and she was legit. I don’t know, and don’t care. Frankly BT, there’s much better ways of doing this!

Windows 7 x64 VT camera II drivers – Cisco Support Community

OK, so it’s horribly hacky, but it DOES work. My camera is working quite happily now!

I ran various tests and found that faxes between ATAs internally were perfectly fine. As soon as they went through the gateways, they failed miserably. I ran through some of Cisco’s help guides but was drawing complete blanks. I’d ran the prserv tool to see what the ATAs were up to whilst making calls. I’d see ‘resync’ go whizzing by with a load of other seemingly random numbers. The word ‘resync’ suggested to me that the ATA was hiccuping on something, and doing something to the audio stream.

Analogue modems expect a constant stream of data. It might get fuzzy, or drop out, but it will always come along in a specific order, at a certain time. It’s predictable. Resyncing something mid-stream isn’t a good idea to a modem. In all cases when I saw the word ‘resync’ the fax would end up corrupted or dropped entirely, depending on when it happened during the call.

I did a little digging on the Cisco TAC case collection, and found what I was looking for. It was something I fiddled with some time ago.

ISDN circuits rely on a clock. Generally speaking, the clock is the telco end. We have four ISDN links on our gateways – two out to the PSTN, and two QSIG links to the old PBX. We had some odd issues with echo, and one thing we tried was forcing clock sync on the E1 controllers.

E1 and T1 controllers can exhibit something known as ‘slipped seconds’. This is basically where the clocks at both ends get slightly out of sync with eachother. In some instances it can cause echo, so we’d nailed up the QSIG links to use the clock at the legacy PBX end. However, it seems with the 2-port WICs, this causes BOTH ports to sync to that clock.

Up to this point it hadn’t been the issue. There was the odd dropped fax, but nothing overly bad. A week or so ago (the week I was off, natch), the faxes all pretty much failed simultaneously. Voice calls remained perfectly fine, which made it all the more perplexing. Luckily I found the info I needed in the TAC collection.

The issue will manifest itself as slipped seconds. On the router, I did the following:

router#sh controller e1 0/0/0
E1 0/0/0 is up.
  Applique type is Channelized E1 - balanced
  No alarms detected.
  alarm-trigger is not set
  Version info Firmware: 20060707, FPGA: 13, spm_count = 0
  Framing is CRC4, Line Code is HDB3, Clock Source is Line.
  CRC Threshold is 320. Reported from firmware  is 320.
  Data in current interval (618 seconds elapsed):
     0 Line Code Violations, 0 Path Code Violations
     181 Slip Secs, 0 Fr Loss Secs, 0 Line Err Secs, 0 Degraded Mins
     181 Errored Secs, 0 Bursty Err Secs, 0 Severely Err Secs, 0 Unavail Secs
  Data in Interval 1:
     0 Line Code Violations, 0 Path Code Violations
     262 Slip Secs, 0 Fr Loss Secs, 0 Line Err Secs, 0 Degraded Mins
     262 Errored Secs, 0 Bursty Err Secs, 0 Severely Err Secs, 0 Unavail Secs

There we see some ‘slip secs’.

The ‘fix’ is quite simple. I switched to the interface, and then issued the commands:

network-clock-participate wic 0
network-clock-select 1 E1 0/0/0

The first line selects the WIC you wish to use, then the second selects the clock source (interface) you wish to use. Once I’d done that, the faxes magically all worked again.

http://www.cisco.com/cisco/web/support/index.html – TAC case collection requires a CCO login.

In my case, I often do ISDN Q.931 debugs to find out what’s going on on my voice gateways. I’ve gotten used to it now, but typing ‘undebug all’ whilst a ton of information whizzes by is less fun than it sounds.

Today, I found this useful command:

“logging synchronous”

If you enable that on your vtys or con (or aux) ports, it basically causes IOS to dump the debug still, but it’ll do it on all other lines EXCEPT the one you’re typing on, leaving it alone. I imagine it could cause some slowdowns as it has to draw your command in the same place, whilst interleaving other junk, so it could potentially cause problems on a direct console, but if you remotely telnet (I really should enable SSH one day), it’ll be a lot easier to see what’s going on.

Here’s the gist of what you need to do:

  Router(config)#line con 0
  Router(config-line)#logging synchronous
  Router(config-line)#line vty 0 4
  Router(config-line)#logging synchronous

Then wr mem as usual.

Easy!