post image

Telenet: flaws by design

"Can I have the Wi-Fi password?" "Sure, what can go wrong?". If you have Telenet at home: A LOT. Each and every Telenet user has a unique login and an email address, the last one is probably known by already everybody. What do you need to take over their network? Just an entry point! That Wi-Fi password usually is no issue to get, you might already have it via Skype.

Now here is the issue: the password forgot feature at "Mijn Telenet" asks only one piece of information: a login/email. As a security check they could have done: phonecall, sms, email?, ask a security question or just ask for a unique client id. What did they chose? The requester's IP. That one IPv4 which is shared by everybody

read more »

post image

Innovate: a remote story.

So we tried Slack, a sentence from their first ad which is actually true. At Innovate we have a team in 3 countries in 4 locations. With in the past even more. Managing to work together when it is impossible to just walk to each other is perfectly possible these days. And Slack helps us a lot with this.

The team

Currently we exist of 4 people. Me, Aaron Gregory, Ethan Gates and Léo Lam. Aaron and Ethan are based in the UK. While me and Léo live on the European mainland in Belgium and France. In the past we had co-workers in several countries like Italy even to New Zealand. This is possible as we also serve a worldwide market, we have customers in almost every continent in the world.

Th

read more »

post image

London can sink, we're fine

"redundancy" 37 matches last month

ITFrame is a vital component of our our infrastructure, it is designed to be the central hub and database of Cast, Control, Apps, Player and DJ. Unfortunately last month(s) was/were not great. Despite the redundancy we had built we had one issue we could not solve. The infrastructure. ITFrame was hosted at OVH Public Cloud, which had several issues to even complete downtime since we switched to it from RunAbove (which has been closed in favor for OVH). This we wanted to solve.

OVH Cloud is something we wanted to avoid in an improved setup, we had too much trouble and no understanding of these. On the path for a new provider we met Linode (where Status

read more »

post image

Aquaris M10 HD Ubuntu Edition review

Warning: I am not a professional journalist. I am a huge fan of Linux (on ARM). This review may not be fully objective. Or contain all correct spelling/grammar.

A few months ago when the Aquaris M10 with Ubuntu was announced I was very excited. First of all the first device with convergence, a dream of Canonical even before Microsoft told a word about "one (scaled down) Windows". But also the first (commercial) tablet with the Linux kernel. And of course an ARM chip! We all know what happened to WinRT, but there is one difference here. It's ubuntu. They have the source of 99% of the packages people use + they have official ports to ARM already available! (I even use those to host this blog)

read more »

post image

Multiarch Docker Images

A few months ago I wrote about Docker and CoreOS on ARM. With the introduction of the C2 by Scaleway we did some changes to that infrastructure. To make it multiarch. That means we both use arm(hf) and x86_64 servers in one cluster, with the same Docker images.
Why should you want this? Some architectures are better suited for some tasks, have a small app that doesn't need much single thread power? Go for ARM. It saves you money, energy and you support innovation. (Probably some other reasons by I am no processor architect)

Get Started with multiarch Docker

Docker was built for x86_64 only. That didn't last long as there began to be ARM ports, with their own special images. But never was

read more »

Why Linode does replace RunAbove

Follow up on my last post. Somebody at OVH decided that 7 days warning is enough to pull the plug at RunAbove (even 0 days for their extra storage disks). 7 days later somebody noticed maybe a migration guide would be handy telling that the VPS SSD replaces the beloved HA S (see last post for why not). Lucky we just had moved all our services away to not OVH.

Ever since my last post my browser had a tab open called Linode Manager. After moving our status page there I kind of fell in love with their service. The support is friendly and direct. I am told that resources are not shared, and my monitoring confirmed that. Sure the manager is not the most recent MEAN powered site, but it works like a charm!

Hmmm...where have I seen this before?

Given guaranteed resources at Linode the pricing reminds me of some other site I had mentioned here.
RunAbove pricing Linode pricing My beloved server for loadbalancers only costs $1 more, yes the RAM is less but overall it is similar (only I still wonder who was thinking logs won't use the 10GB storage...). I have to say the pricing is very nice for the setup we want, which was the reason why we liked RunAbove in the past.

Move the stuff!

Past days we finalized the move of ITFrame after 3 months with OVH Cloud. The ITFrame servers are becoming more important to us and we can't afford yet another downtime caused by the hypervisor or an unexplainable slowness due to shared CPU power. Issues we had experienced at least once a month.

Lindode offers 2 locations in Europe: London and Frankfurt. The downtime of a whole datacenter can happen so we need to work that around. Yes with OVH we could have done the same only often the same issue happened at the other location too. Not to talk about the SPOF that seems to be able to cut OVH off the network for whole Europe.
The average ping between these two Linode locations was 11ms, perfect to setup MongoDB replication. We went further and set up a loadbalancer, worker node and database server in both locations. These are all linked to each other to fall back on the server on the other side of the channel in case it is needed. Currently ITFrame shouldn't suffer from any infrastructure issues anymore, in the worst case another datacenter will take over all the traffic :D

Please it is 2016

IPv6. Yes, you already know that I will say. RunAbove did never support IPv6 even though dedicated servers and the old VPS offer did. Okay no problem, you have other priorities. With OVH Cloud they rebuilt the whole infrastructure (hopefully learning from past mistakes) the perfect chance to implement full IPv6. Nope. Still no word on it, no wonder nobody uses it.

Linode on the other side. They don't use OpenStack like OVH (which btw has full IPv6 out of the box), they made their own. Meaning they spent time on adding IPv6. And yes they have it! Finally we can offer IPv6 to our clients which allows us to move on to the future.

tl;dr

Linode is better. We moved all our cloud x86 servers over.

Why OVH Cloud doesn't replace RunAbove

I am a big fan of OVH. They got a great network, best DDoS protection, very good pricing, nice hardware but they made a mistake with their cloud.
More than a year ago we at Innovate Technologies migrated many of our services to the new cloud service under the sub-diary of OVH called RunAbove.

They had a nice offering. Sandboxes, with SSD, 2 or 4GB memory and shared resources. Great to develop on or run a small unimportant server. Out of benchmarks it was noticeable that not every sandbox had the same performance and some went slow for a period of time. For $2.5 and $5 not a bad offer.
The 2nd series was the (internally called) HA ones, with dedicated resources(!) and distributed storage for a quick recovery. Starting at $9. This offer was great for smaller servers that needed performance and availability. Even more perfect for a load balancer redirecting traffic to the 3rd option which was just a whole dedicated server with local SSD.

In preparation for the OVH Summit RunAbove's classic offer as described above was stopped to rebrand it as a Lab for new services. Old customers can keep their servers but are suggested to move to OVH Cloud (some support docs seems to force it very hard), an "improved" copy of RunAbove on their premium brand OVH.com.
First of all, you can't simply move the data, IP, settings etc. We had over 50 servers at one point on RunAbove. You just don't move those in 123. Sure they don't force us to move but keeping a large set of servers or even expanding them on a service that gets treated as a beta lab is a bad idea.
But the reason we didn't move all servers was simple. OVH Cloud has two offers. A sandbox like server called VPS SSD, as said above our experience was not very good especially if you want dedicated performance. And a ram/cpu "dedicated" server offer which carry the names of their dedicated server plans. Starting at 30 euros that is 3 times more than the $9 of RunAbove. For a small loadbalancer this is maybe overkill if you're used to the RA offer. And you don't want a vital server to run on shared resources. You don't want your loadbalancer slowed down due a neighbor being a CI that compiles Linux kernels all day.

For over a few months I gave OVH Cloud a try, benchmarks showed promises but benchmarks are just numbers. Over the past months I had several times frozen VMs or a hypervisor that was down. For a VPS SSD that means complete downtime, since it is our loadbalancer... You can guess it. The past weeks our performance of the ITFrame system went down. Not sure where this could be I delayed investigation as there was more important things to do. At yet another downtime I was sick of it and set up a 2nd cluster (good I had backups) at 2 in the night. Suddenly I noticed Nginx started a lot faster (before it took like a minute, large config files? bug? We use 1.9.x after all). Performance of only the load balancer was a lot faster (5-7x). Why? Oversold shared resources. Something RunAbove's HA did not do.

For being a big OVH fan and using only their servers for over 2 years (not any longer) this is very disappointing.
For the ITFrame cluster? I sit and wait for Scaleway's x86 offer (I saw the code, it is coming!) to build a hybrid cluster of ARM and x86 for the Node.js workers. In meantime. Pray and look for other providers.