Studying up for my BGP+MPLS exam as part of the CCIP track and was working out some labs on BGP Confederations. We don’t use them at work so they were something that I’ve never experienced. First of all I can’t figure out why anyone would want to use them over route reflectors. A couple things I’ve learned:
I had been wanting to get BackTrack up and running under ESXi and the fastest way is to download the prebuilt Vmware image but it didn’t work:
Message from server: Failed to open disk scsi0:0: Unsupported and/or invalid disk type 7. Did you forget to import the disk first?Unable to create virtual SCSI device for scsi0:0, '/vmfs/volumes/f9f3a780-a066e304/Backtrack4-Final_ORIG/BackTrack4-Final.vmdk' Module DevicePowerOn power on failed. info 23.03.2010 22:05:40
After a bunch of trouble shooting, and finding nothing I decided to import it via VMware vCenter Converter. Now I was able to get BackTrack4 to boot but /dev wouldn’t populate and I got dumped into a busybox shell. Once I changed the SCSI controller from Paravirtualized to BusLogic Parallel, BackTrack4 booted perfectly. Logged in and started X.
Works pretty slick so far.
For all of us that run our own home labs we know (or should know) the price of what it costs us to run our equipment every month. Previously I spent some time with a Kill-A-Watt meter performing an energy audit on my equipment and found that it uses about 650 watt/hrs of power. At 12 cents per kwh I’m spending $78/month on my lab, over $900/year. 60% of our monthly electric bill goes towards my work hobby.
I’ve tried to slim some servers down by moving them to a laptop running Debian & VMware Server but there are creep-cpu-usage problems involved with that. The laptop running VMware Server wasn’t all the beefy either, a Athlon 2800+ with 512mb of RAM. 1 or 2 guest OSs before things start to slow down. No RAID setup, no firewire, no expandability; a good stop-gap system until I could figure out what I wanted to do.
Through a lot of reading I’ve decided that an VMware ESXi system was the way to go if I wanted to virtualize my servers and have enough room left over to delve into Cisco Call-Managers, IDS/IPS systems, Nessus scanners, Asterisk, Olive’s, and just stand alone systems for trying new things. I also needed a NAS to centralize all the usb/firewire drives that I have laying about. The NAS needed to have RAID capabilities and act as an iSCSI target so I’ve chosen to go with FreeNAS. Read more…
To redistribute between MP-BGP (Core) and OSPF (CPE), you must redistribute MP-BGP into OSPF and OSPF into MP-BGP.
- Working MPLS Core
- Established BGP neighbors between PE devices
In the PE router, a VRF specific OSPF process needs to be started:
PE01(config)#router ospf 200 vrf Cust_B
PE01(config-router)#redistribute bgp 65000 subnets
Very important to include the subnets option unless you are and will ONLY work with classful networks. You will get a warning that only classful networks will be redistributed if you don’t include it. Chances of only working with classful networks are slim (no use of loopbacks), so probably best practice to always include ’subnets’. In my study material I was told that having a customer use OSPF as their routing protocol was not a good idea as the routers were limited to 32 OSPF processes (28 usable). Just checked Cisco and looks like that has been fixed with IOS 12.3(4)T and higher.
PE01(config-router)#router bgp 65000
PE01(config-router)#address-family ipv4 vrf Cust_B
PE01(config-router-af)redistribute ospf 200 vrf Cust_B
Pretty easy setup, just need to remember the redistribution is bi-directional.
One of the steps I needed to take to virtualize servers was to setup vlan trunking between a Linux server and my switch and then have VMware Server bridge to those vlan interfaces.
Here are the steps involved:
While in the process of pre-staging a consolidation of my servers to one hefty VMware server( instead of multiple machines idling at 80 to 250 watts each) I realized that I could begin the building, patching and installing process on my laptop because one of the coolest features of VMware is that you can move the guest OS from computer to computer and power them up, they never know the difference.
In starting the build process I ran into a snag with Debian Lenny. The symlinked gcc version is 4.3 and VMware Tools was complaining that since the kernel was compiled with 4.1, I should use that instead. It suggested I create a variable and assign it the path to gcc-4.1. Ran into some problems doing this and eventually discovered that sudo strips most environmental variables before it executes the program you want to run. So the variable ‘CC’ was getting removed before the vmware tools script was run.
- Include changing the symlink for /usr/bin/gcc to /usr/bin/gcc-4.1 for the build process and then back to 4.2 (Quick Fix)
- Going into root to add the variable and then installing the (Easiest)
- Adding ‘CC’ to the allowable variables in sudo and/or disabling ‘env_reset’ in the options (Opens potential holes for users to exploit).
I also found that a lot of people were having the same issue as me and posted a reply on another blog about what I found
In using Dynamips while now while I study for my CCNP and discovered that even though with a good idle_pc value, the CPU utilisation spikes after the con session times-out. I initially noticed that once the router boots and you’re prompted ‘Press RETURN to begin!’ and the interfaces normalise, the CPU stays pretty high until you press RETURN to begin the console session. Same goes for when the console session times out.
Running 6 2600s at ~30% processor usage and having one session time out brings the CPU usage up to 90% until I press enter again to clear the ‘you’ve timed-out’ message. If several time out, the overall load goes up pretty quick, 7+ at times.
The key is to just set the exec-timeout to 0 -
Router(config)#line con 0
Router(config-line)#exec-time 0 0
With avoiding the time-outs, I should be able to load twice the number of routers now.
I’ve been able to trouble shoot and revamp CGI Perl code in my job that other people have written but never sat down and learned Perl well enough to write my own. Programming isn’t new to me. My first code was Basic like most peoples. Either drawing pixel images on my Commodore or simple text games on an Apple IIe. I’ve dabbled in BASH scripts a bit, mostly for parsing data or performing loop functions.
We had went out for dinner one night, and I had something with cheese in it, or some fries with ranch dipping sauce. I took my normal 3 lactaid pills to help stave off most of the discomfort dairy has given me over the past 10 years or so, drank a pitcher of Guinness (read some where that it helped digest dairy) and felt no ill affects an hour or two later. Lactaid has never worked that well. It helped prevent some of the more unpleasant problems but always fell short of stopping all symptoms.
I had put my relay rack over the summer and slowly added parts to it. When I had to rewire the Cisco devices for study purposes or because I needed to change a computer out it was done with little regard to neatness (ie. sloppy is fast). I went from using the Costco special baker’s rack:
To a 7′ Chatsworth (that I got for free):