Profile cover photo
Profile photo
Hugues Malphettes
172 followers
172 followers
About
Hugues's posts

Post has attachment
Yoda Raphaelle and Chewbalouise
Photo
Photo
18/07/2014
2 Photos - View album

Post has attachment
Germany wins 
Photo

Post has attachment
Super Raphaelle and Louise the butterfly
Photo

Post has attachment
Raphaelle on the keyboard
Photo

Post has attachment

Post has attachment
I am presenting a short oboe recital this Wednesday at Orchard Road Presbyterian Church at 12:30.

Please join me at "Sanctuary at Orchard" for a peaceful break in the middle of your busy Singaporean day.

The multi talented Jens Kludge will accompany me on the organ for a Fantasy in F minor by J.L. Krebs and on the piano for an Italian Dance by M. Dring.

Veda Lim is my awesome coach.
Photo

Post has attachment
9 months 5 teeth and her first pair of shoes.
Photo

Post has attachment
Good morning latte
Thanks Stanly!
Photo

Post has attachment
Docker.io: PaaS is getting upgraded to IaaS.

Virtualisation of your processes without the weight of a full Virtual Machine. Root access to the OS where your apps are running without a IaaS.

The Platform as a Service fundamental brick - the container for the app execution - is getting mainstream with an opensource implementation:

    http://www.docker.io/

"hello world" with this container is as fast as executing 'console.log("hello world");' in your browser.
It runs in an isolated linux of your choice with its own packages, network interface, mix of volatile and persistent file-system: you have root access.

The 5 minutes video on the homepage of docker.io left me awestruck.

No more distinction between apps, services, one-time tasks:
node.js, python, ruby, ssh-server, redis, postgres, couchdb are the first 5 examples documented:
http://docs.docker.io/en/latest/

The reluctance of the Cloud Foundry team to make it easy to run Elasticsearch on Cloudfoundry is disappointing:
http://support.cloudfoundry.com/entries/24469701--ulimit-n-how-to-increase-the-limit-of-files-opened-above-16k

I am convinced that there is room for more flexibility.
I don't want to run my own infrastructure.

Elasticsearch is consumed as a database but it runs as an application.

Say we find a PaaS or a IaaS that happily runs Docker's images.
We would run:
- an ssh daemon for debugging
- our nodejs apps with all the native libs we want: it is our OS after all.
- our Elasticsearch can run as a cluster because opening multiple ports is trivial; we can attach a persistent file system too: Amazon's EBS is right there.

And we would run that very same Docker image on an linux VM for development purpose on our laptop.

docker.io is standing on the shoulders of giants. Thanks to them we get 90% of a IaaS at the price of running an isolated unix process.

Kudos to dotcloud for brilliantly decoupling this component out of their PaaS and open-sourcing it.

So next time I need to run something on a IaaS, I'll most likely run it through Docker.io.
Those days are counted though: as soon as a PaaS hosts Docker images for us we will jump there and never look back.

Post has attachment
Elasticsearch indexes backup and clones with S3.

At Stoic we are currently using Elasticsearch as our primary datastore.
It is either amazing or good enough.

In the area of good-enough; here is how we tackle durability:
- Setup the S3 Gateway: it will periodically send the indexes and cluster state into an S3 bucket.
- Copy the S3 bucket somewhere else.

Here is a short recipe to clone the data from S3 to a local setup of Elasticsearch
- Setup the same version of Elasticsearch that then one we use: https://github.com/hmalphettes/elasticsearch-cloudfoundry-packager
- Download the content of the backup: ${bucket-name}/${elasticsearch-cluster-name}
- Copy this folder into elasticsearch/works/${elasticsearch-cluster-name}
- Start Elasticsearch with the same clustername and with the Shared Filesystem Gateway: ES_CLUSTER_NAME=${elasticsearch-cluster-name} GATEWAY_TYPE=fs

Enjoy debugging on your local machine.
Wait while more posts are being loaded