At work I always prefer KVM hosts for reasons such as flexible, free and GUI-less. Yet I never bothered to go deeper into the networking features of Libvirt, so I only connect VMs to the host networks via Linux Bridges or OvS. Far far away from fancy virtual libvirt networks.

Even with this simple networking approach I recently faced a tedious task of reconnecting VMs to different bridges on-the-fly.
My use case came from a need to connect a single traffic generator VM to the different access ports of virtual CPEs. Essentially this meant that I need to reconnect my traffic generator interfaces to different bridges back and forth:

Apparently there is no such virsh command that will allow you to change bridge attachments for networking devices, so a bit of bash-ing came just handy.

xrdp is defacto the default RDP server for Linux systems sharing with VNC the remote access solution olympus. I personally found it more resource friendly and feature rich compared to VNC solutions I tried. The only problem I found with xrdp is that current Ubuntu LTS release Xenial 16.04 has a way outdated 0.6.1-2 version of xrdp in the packages repo. This version has no shared clipboard support, which makes remote support/remote access a tedious task.

While Amazon Linux AMI has yum as a package manager, it is not that all compatible with any RHEL or CentOS distributive. A lot of changes that AWS team brought into this image made it a separate distro, so no eyebrows should be given when battle-tested procedure to install python3 will fail on Amazon Linux. (Yeah, python3 does not come included yet in Amazon Linux)

I wrote about the Yang Explorer in a docker quite some time ago, Yang Explorer was v0.6 at that time. Back then the motivation to create a docker image was pretty simple – installation was a pain in v0.6, it is still a pain, but the official version bumped to 0.8(beta).

So I decided to re-build an image, now using Alpine Linux as a base image to reduce the size.

Hugo gets a lot of attention these days, it is basically snapping at the Jekyll' heels which is still the king of the hill! I don’t know if Hugo' popularity coupled with the fastest static-site-generator statement, but for me “speed” is not the issue at all. A personal blog normally has few hundreds posts, not even close to thousands to be worried about slowness.

Then if it is not for speed then why did I choose Hugo? Because it became a solid product with a crowded community and all the common features available. (To be honest I also got an illusion that one day I might start sharpen my Go skills through Hugo as well).

As you already noticed, this blog is powered by Hugo, is running on GitLab pages, with SSL certificate from CloudFlare and costs me $0. And I would like to write down the key pieces that’ll probably be of help on your path to a zero-cost personal blog/cv/landing/etc.

I started to play with Go aka Golang. Yeah, leaving the comfort zone, all that buzz. And for quite some time I’ve been engaged with VS Code whenever/wherever I did dev activities. VS Code has a solid Go support via its official extension: This extension adds rich language support for the Go language to VS Code, including: Completion Lists (using gocode) Signature Help (using gogetdoc or godef+godoc) Quick Info (using gogetdoc or godef+godoc) Goto Definition (using gogetdoc or godef+godoc) Find References (using guru) File outline (using go-outline) Workspace symbol search (using go-symbols) Rename (using gorename) Build-on-save (using go build and go test) Lint-on-save (using golint or gometalinter) Format (using goreturns or goimports or gofmt) Generate unit tests skeleton (using gotests) Add Imports (using gopkgs) Add/Remove Tags on struct fields (using gomodifytags) Semantic/Syntactic error reporting as you type (using gotype-live) Mark that gotools in the brackets, these ones are powering all that extra functionality and got installed into your GOPATH once you install them via VS Code.
Nothing bad in knowing how many lines of code or text out there in your repo. You don’t even need your VCS to convey this analytics. All you need is git, grep and wc. 1 2 3 4 5 6 7 8 9 10 # count lines in .py and .robot files in /nuage-cats dir of the repo $ git ls-files nuage-cats/ | grep -E ".*(py|robot)" | xargs wc -l 0 nuage-cats/robot_lib/__init__.

Cloud-native revolution pointed out the fact that the microservice is the new building block and your best friends now are Containers, AWS, GCE, Openshift, Kubernetes, you-name-it. But suddenly micro became not that granular enough and people started talking about serverless functions!

When I decided to step in the serverless property I chose AWS Lambda as my instrument of choice. As for experimental subject, I picked up one of my existing projects - a script that tracks new documentation releases for Nokia IP/SDN products (which I aggregate at

Given that not so many posts are going deeper than onboarding a simplest function, I decided to write down the key pieces I needed to uncover to push a real code to the Lambda.

Buckle up, our agenda is fascinating:

  • testing basic Lambda onboarding process powered by Serverless framework
  • accessing files in AWS S3 from within our Lambda with boto3 package and custom AWS IAM role
  • packaging non-standard python modules for our Lambda
  • exploring ways to provision shared code for Lambdas
  • and using path variables to branch out the code in Lambda

virsh is a goto console utility for managing Qemu/KVM virtual machines. But when it comes to deletion of the VMs you better keep calm - there is no single command to destroy the VM, its definition XML file and disk image.

Probably not a big problem if you have a long-living VMs, but if you in a testing environment it is naturally to spawn and kill VMs quite often. Lets see how xargs can help us with that routine.

It may very well be that VPLS days are numbered and EVPN is to blame. Nevertheless, it would be naive to expect VPLS extinction in the near future. With all its shortcomings VPLS is still very well standardized, interop-proven and has a huge footprint in MPLS networks of various scale.

In this post I will cover theory and configuration parts for one particular flavor of VPLS signalling - BGP VPLS (aka Kompella VPLS) defined in RFC4761. I’ll start with simple single home VPLS scenario while multi-homing techniques and some advanced configurations might appear in a separate post later.

In this topic the following SW releases were used:

  • Nokia (Alcatel-Lucent) VSR 14.0.R4
  • Juniper vMX 14.1R1.10