HNHacker News
TopNewBestAskShowJobs

ermintrude

42 karma · joined November 22, 2013

submissionscomments
ermintrude··on What if you didn't need money or attention?
If you pick up almost any book on eastern philosophy you'll find that living in the present is the key. Give up the future. It won't be how you imagine anyway.
ermintrude··on What if you didn't need money or attention?
Wrong. Your biggest problem is believing you have to improve yourself.
ermintrude··on Just what we need. Another package manager.
Have you ever worked with other people who use different OSes, or need different versions of libraries installed?
ermintrude··on Squirt.io – Readability Meets Spritz Speed Reading
Why does each user need to be given an ID? How private is this?
ermintrude··on Python vs Julia – an example from machine learning
Can you also call C++ libraries directly from Julia?
ermintrude··on What Gordon Ramsay Taught Me About American Businessmen
> Or is it just to stroke my ego*

> *And I realize that's ironic coming from a guy who just wrote an 800-word blog post telling you how to think.

It's not ironic.

ermintrude··on Flask API – Browsable Web APIs for Flask
I think the architecture supports pluggable backends, but AFAIK mongo is the best supported (and is the only one I know much about). There is a backend for elasticsearch (https://github.com/petrjasek/eve-elastic) but I've never used it and don't know how complete it is. From looking on github support for SQLAlchemy is slowly in the works.
ermintrude··on Flask API – Browsable Web APIs for Flask
> Is there a special "API key" for your web app?

And if there is, what's the point? Would the only reason to ever change that key be to force old clients to upgrade if the API changes?

ermintrude··on Flask API – Browsable Web APIs for Flask
You should check out eve (http://python-eve.org) - it's amazing for rapid prototyping (if the fact it requires Mongo isn't an issue).

I'm building an AngularJS + REST server using Eve for the server. It has saved me so much time that I'm putting my dev effort mainly into a comprehensive test suite because there's so little dev work to do (btw - I'm not the author, it's just the fastest framework I've found for developing REST APIs - and I've tried Django REST framework & tastypie).

ermintrude··on Flask API – Browsable Web APIs for Flask
Maybe you could contribute to Eve (http://python-eve.org) instead (which is awesome) and already includes some of the features on your roadmap.
ermintrude··on Cluster-Level Container Deployment with Fleet
I think the big gap with docker is logging. If you're going down the coreos route you'd probably want your logs centralising somewhere. That means that you need to configure your containers to run something like supervisord and rsyslog to log to a remote server which is a bit of a hassle, especially if you were making use of public images since you'll need to add all this boilerplate in yourself and rebuild the images.

Alternatively I guess you could log to the host if there are config options to log directly to syslog and you could then ship them from there.

I spent a couple of weeks writing a python app and the same again trying to sort out a flexible logging system. Ah well, I guess it's still under development, but it'd be great if they prioritised this since it's one of the major things stopping me using docker (and obviously I'm so important they should reprioritise around me!)

I really wish coreos was bundled with a package manager though. Not having telnet or the ability to apt-get install things easily makes any sort of debugging and real-world use infeasible for me... :-(

ermintrude··on Docker at Spotify
It seems cpu.shares allow a proportional amount of CPU to be allocated, but LXC may use up to 100% of all idle CPUs. You can't specify a hard-limit of "only use 1GHz when I have a 2GHz CPU" apparently (according to here: http://comments.gmane.org/gmane.linux.kernel.containers.lxc....). So I guess that could mean that if the containers are using 100% of all CPUs, there wouldn't be resources left e.g. for monitoring on the host? Anyone got any experience with this?
ermintrude··on Docker at Spotify
Thanks. I had seen packer.io but then forgot about it.

I had another look at Docker in November and had issues installing what I needed via apt for some reason. It wasn't a drop-in replacement for VMs for several reasons (I wanted an alternative backend to provision with salt), but when it hits 1.0 I'll have a look again, if not before.

ermintrude··on Docker at Spotify
They let devs ssh onto production boxes and do what they want?!?! "Things diverge pretty quickly" - I'm sure they do. That sounds like a recipe for disaster...

Docker looks promising but until there's a way to allocate maximum resources to a container I wouldn't use it in production. VMs are much slower to start but at least a runaway process on one won't affect other VMs on the same hardware.

Docker only lets you hint that only a certain number of cores should be used, so a bad process might monopolise the physical machine. Also until other tools (ansible, salt, chef, etc) help with provisioning you need to edit the dockerfile to change parameters (eg for settings that depend on your environment like smtp end points test vs prod, etc).

If these points are addressed I think it'll be awesome though.

ermintrude··on Ansible 1.4 released
I'm working on a large project running on AWS. We're building an n-tier architecture with many roles in auto-scaling groups.

We wasted the first couple of months writing wrappers around SSH to let us provision servers with salt, which if we'd been using ansible would have been completely unnecessary. We did evaluate ansible at the start of the year, but at that time it didn't look quite as mature so we went with salt.

I've used ansible myself and it's far simpler than salt, and quicker to get started with. Admittedly I haven't used it to deploy the same complex software as we are doing with salt, where orchestration is fiddly. Overall, the thing that lets salt down is its appalling documentation, poor support (the IRC channel has been practically useless) and complex terminology (what's the difference between a module and a state? Who knows...).

Salt does have one thing in its favour though - the ability to run things in parallel. For example a new node can request itself to be configured which is very useful when running in auto-scaling groups. I think this is coming in ansible, but it wasn't at the same level as in salt last I checked.

With salt, when a new node comes up (in an auto-scaling group for example), it can request itself to be configured. Once finished it can fire an event to the saltmaster which can trigger it to reconfigure other nodes. This, for example, means when a new web server comes up, the app can be deployed to it and the node can then instruct the saltmaster to update the load balancer configs to send traffic to it. This event system is powerful because it lets multiple tasks take place in parallel, but it is complicated.

If ansible had an intuitive way of running tasks in parallel (with locks through critical sections, the ability to target a subset of matched nodes (to back up a single server in a pool for example), etc.), then I think it'd beat salt hands down.

For the majority of cases I'd definitely recommend ansible over salt. And as for chef/puppet after using them both on earlier projects, we started looking for something else that could do orchestrations as well as provisioning and that had saner authentication when running in the cloud.

Good work ansible team!

ermintrude··on Ansible 1.4 released
Personally I think the ansible docs are very good. If you think they're bad, try looking at the salt docs which are absolutely rubbish.
← PreviousPage 2 of 2