Outside the Asylum

Author: daniel@shirow.systems

  • Cisco router lab

    It can be very hard to get hands on practise working with networking technologies. It usually requires a stack of routers and switches to be able to configure up a meaningful topology. There are simulator software packages but they don’t run the code that runs on real routers.

    Recently I discovered dynamips written by Christophe Fillot who is either a phd student or faculty member at UTC (University of Technology of Compiegne, France), I’m not sure which. Its an application that emulates the MIPS hardware of a range of Cisco routers. Cisco IOS software can be booted by this emulator. IOS software is not distributed with dynamips as that would be a violation of Cisco’s copyright. If you have a Cisco support contract you will have access to IOS software downloads. In addition to the core dynamips application there is dynagen (written by a number of people) which is a python user interface which makes interacting with dynamips a whole lot easier.

    To get started, head over to the tutorial on the dynagen website

  • APC vs Cell site. Tank 1, Site 0

    there-has-been-significant-damage-at-various-sites.jpg
    Engadget Post

    Pot sure brings out the paranoia in people. Ex-phone technician/mechanic steals a restored APC and goes on a rampage ramming cell sites and power substations.

  • Routing redistribution

    I had a problem a little while ago with a customer who was using RIP between us and them. We’re redistributing RIP routes into BGP to be passed along internally. The problem was, a routing advertisement loop manifested itself. Dagnabbit! Took me a while to figure out what was going on. The symptom was that only about 10-20% of pings were succeeding. I then tracked it down to a route that was coming and going, its metric was winding up to 16 (RIP limit) then being discarded. To cut a long story short, what was happening was; A) a BGP route was advertised to one AS (lets call it 1 ) from another (lets call it 2). B) this route was then redist’d into RIP (for the customer device). C) another router in AS1 has an interface in the same subnet as the RIP advertisement. This router picks up the RIP route and dutifully pops it into BGP. D) the first router gets the route via BGP from the second router and there is the loop.

    drawing1.png

    I had to figure out how to stop the route going around in the loop. In the end i settled on making RIP unicast. I put neighbor statements into the RIP config for each CPE device. This seemed to fix the problem. Later i discovered that i could filter the routes being advertised and learned, either based on tags applied when redistributing or based on a list of networks. I’m not sure what the best way is or the pros and cons of each. The RIP neighbor statements seem to be the easiest, just one line of config! I’m just worried I’ve missed something or there’s some gotcha. The tagging of routes when they’re redistributed from BGP to RIP and then filtering when redistributing from RIP to BGP seems to be the most comprehensive fix, but it is a lot of config. If unicast RIP is just as effective, then I’d rather use that. I like simple fixes. Comments?

  • Confluence – enterprise wiki

    confluence_logo.gif In a corporate work environment a wiki is such a fantastic tool for collaboration. I’ve tried out a lot and none were quite good enough for enterprise use. Then a friend put me onto Confluence. Its easy to setup, backup, upgrade and migrate. I’m a fan now. I went to a talk last week at Atlassian’s offices here in Sydney on how to promote wiki use in your organisation. The basis for the talk was the site wikipatterns.com. Their new office is very nice and they provided beer and pizza; quality pizza and my favorite beer, Stella Artois. awesome!

    confluence-model.gifI even use an instance of confluence for my personal stuff; both general and career related. The main instance I run on my laptop. Every so often i back it up (easy, just a couple of clicks and save a zip file to a usb memory stick) and take it home to restore onto Confluence instances i have running there. I have 3 instances installed; on OSX, Solaris and Windows. All running on Tomcat and MySQL.

    Highly unlikely that all of them could be lost. Saaaaaafe.

  • my ideal home workstation

    macprolarge01_20060807.jpgrr2322_card_big.jpg
    e8_back_g.gife8_frontopen_g.gif

    I’ve been pondering what I’d set up at home if I was going to go all out but keep the cost within realistic limits.

    • Mac Pro dual quad core Xeon, lots of RAM
    • Highpoint RocketRAID 2322 PCI-Express x4 SATA II RAID Controller w/ RAID 5 and External Infiniband Support or maybe a pair of them.
    • Enhance E8-ML Multi-Lane 8-bay SATA Enclosure (with Two Mini-SAS to Infiniband 4-Channel MultiLane Cables)
    • Eight 750GB SATA-II HDDs or whatever’s biggest.

    UPDATE (11-04-2021): it’s now ~15years later. I’ve just upgraded the home lab to 10GE and arrays of 12 and 14TB disks. NVMe SSDs are up at 2 and 4TB in size, with speeds and latencies that weren’t even dreamed of not long ago. Latest lab server is an AMD 8core/16thread cpu.

  • ZFS and Containers

    sc_fig_11.gif

    Upside: good granular resource management. efficient disk usage.
    Downside: one Solaris kernel. shared ethernet interfaces.

    Solaris Containers ( = Zones plus Resource Management) and ZFS is pure joy, especially the ability to clone zones to save disk space. There is a bit of a gotcha though. As Solaris hasn’t got boot support for ZFS yet, there is an issue in that you cant upgrade a zone if its root is sitting on a ZFS volume. I hear that the next release of Solaris (07/07?) will have ZFS boot support and will fix this issue. Solaris Internals has lots of great info.

  • Ruby enlightenment

    I’d heard about Ruby when the whole Ruby on Rails hype wave hit a few years ago. For whatever reason I didn’t give it a proper look until recently. Wow. I’m a changed person. For a long time I’d considered coding to be not my thing; too tedious and boring. I’m an pragmatic engineer type. I like to just get things done.

    Starting to read the books on Ruby, it feels like the language was written for people like me. For the first time I actually enjoy writing code and learning how to do things in better ways.

    Then there’s Ruby on Rails, a web framework written in Ruby. I cant believe how easy it is to get started.

    Shakabuku!

    Debi: You know what you need?
    Marty: What?
    Debi: Shakabuku.
    Marty: You wanna tell me what that means?
    Debi: It’s a swift, spiritual kick to the head that alters your reality forever.
    Marty: Oh, that’d be good. I think.

  • Sun Multithreaded 10 GbE Networking Card

    8178.jpeg

    Sweet!

    Maximum Ethernet transfer rate
    • 10 Gb/sec for each single port
    • 16 Gb/sec total for both ports
    Host Interface x8 lane PCI-Express 1.1
    Network Interface Two instantiations of XAUI, IEEE 802.ae, 2002 compliant
    Optics 10GBase-SR, 10GBase-LR
    IEEE 802.3ae 2002 compliant

  • Sun LDOMs on Coolthreads servers

     ig_ldoms_howitworks.gif

    I set up LDOMs on a T2000 recently to have a play and learn about them. This pdf is a good intro to LDOMs.

    Basically, the steps are; install solaris (11/06 or greater), upgrade firmware, install LDOM software, configure installed solaris instance as the first control LDOM, reboot. repeat, rinse, RTFM etc.

    A word of advice, always read release notes. I hit a bug where I could not add/remove cpus from the LDOMs I had created. big problem! After bit of googling I found the section “Dynamic Reconfiguration of Virtual CPUs with Cryptographic Units” in the release notes. Bingo.

    So I feel that LDOMs are a bit bleeding edge still. I’ll wait a bit before i start using them.