I left my server in El Segundo
If you're reading this, then you're looking at the new home of djangoproject.com.
Django's adoption has skyrocketed over the last six months or so, and with the added popularity has come increasing traffic to this site — we're doing close to eight million hits each month, and growing. While the increased interest couldn't make us happier, it made the old, cranky server running the site incredibly upset. Over the past weeks we've been looking around for a new home, and I'm glad to say we found it:
About a month ago, Media Temple generously offered to take up hosting our site. They've given us a screamingly fast dedicated server, free hosting, and wonderful support. Even better, this is only Media Temple's first step into the wonderful world of Django; they've got some very cool Django hosting plans coming in the near future.
We're very excited about the things we'll be able to do with our new, more powerful server. In fact, we've already started: this weekend, Matt Croydon and Joe Heck set up the Django Buildbot which has already helped us catch a few nasty bugs. We're really looking forward to finding other areas in which our newfound power can help us develop Django faster and better.
So, once again, a big thanks to Media Temple — and especially Chris Lea — for their support.
Bonus points to anyone who gets the reference for the title...
Posted by Jacob on September 17, 2007
Comments
Vance Dubberly September 17, 2007 at 7:39 p.m.
Well this is promising I've been listening to Media Temple promise django support on their GS for a year. While they've still not gotten off their duffs supporting hosting applications it's good to see they are supporting the project. :)
Hippety-Hop September 17, 2007 at 9:30 p.m.
Many thanks to MediaTemple on behalf of the community.
I've been very happy with them in the past... now if they'd only get off of CentOS (or at least upgrade from 4.4) so having up to date software would be easier, I'd be back in a heartbeat.
Still, they're a great company, and it's great to have their support.
John September 17, 2007 at 9:30 p.m.
MUCH FASTER! Of course I live in LA so it should be, but thanks to MediaTemple
Matt Croydon September 17, 2007 at 9:46 p.m.
I'm shocked that the Wikipedia page doesn't reference the (fantastic) Fatboy Slim vampire mix of the song alluded to in the title.
Maybe we can spare a line or two from the list of destroyed Star Wars planets article...
rit lim September 17, 2007 at 9:49 p.m.
The link in "Hosting graciously provided by (mt)" is wrong. It should mediatemple.net, not mediatemplate.net
Jacob September 17, 2007 at 10:31 p.m.
Thanks, rit lim -- my fingers must be hard-wired to "template" :)
Jeff Croft September 17, 2007 at 11:29 p.m.
Best. Title. Evar! (for a blog post).
Roland September 18, 2007 at 2:05 a.m.
Trac seem to have a problem. Maybe a result of the move ?
Oops…
Trac detected an internal error:
If you think this really should work and you can reproduce it, you should consider reporting this problem to the Trac team.
Go to http://trac.edgewall.org/ and create a new ticket where you describe the problem, how to reproduce it. Don't forget to include the Python traceback found below.
TracGuide — The Trac User and Administration Guide
Python Traceback
Traceback (most recent call last):
File "/home/trac/trac/web/main.py", line 406, in dispatch_request
dispatcher.dispatch(req)
File "/home/trac/trac/web/main.py", line 191, in dispatch
chosen_handler = self._pre_process_request(req, chosen_handler)
File "/home/trac/trac/web/main.py", line 263, in _pre_process_request
chosen_handler = f.pre_process_request(req, chosen_handler)
File "/home/trac/trac/versioncontrol/api.py", line 73, in pre_process_request
self.get_repository(req.authname).sync()
File "/home/trac/trac/versioncontrol/api.py", line 104, in get_repository
repos = self._connector.get_repository(rtype, rdir, authname)
File "/home/trac/trac/versioncontrol/svn_fs.py", line 234, in get_repository
repos = SubversionRepository(dir, None, self.log)
File "/home/trac/trac/versioncontrol/svn_fs.py", line 256, in __init__
self.pool = Pool()
File "/home/trac/trac/versioncontrol/svn_fs.py", line 133, in __init__
self._pool = core.svn_pool_create(self._parent_pool())
File "/var/lib/python-support/python2.5/svn/core.py", line 177, in svn_pool_create
return Pool(parent_pool)
File "/var/lib/python-support/python2.5/libsvn/core.py", line 1082, in svn_pool_create
return apply(_core.svn_pool_create, args)
File "/var/lib/python-support/python2.5/libsvn/core.py", line 3105, in _wrap
obj.set_parent_pool(self)
File "/var/lib/python-support/python2.5/libsvn/core.py", line 2996, in set_parent_pool
self._parent_pool = parent_pool or application_pool
File "/var/lib/python-support/python2.5/libsvn/core.py", line 2984, in <lambda>
__setattr__ = lambda self, name, value: _swig_setattr(self, apr_pool_t, name, value)
File "/var/lib/python-support/python2.5/libsvn/core.py", line 24, in _swig_setattr
return _swig_setattr_nondynamic(self,class_type,name,value,0)
File "/var/lib/python-support/python2.5/libsvn/core.py", line 19, in _swig_setattr_nondynamic
self.__dict__[name] = value
RuntimeError: instance.__dict__ not accessible in restricted mode
vicent roca September 18, 2007 at 3:41 a.m.
Hi!
Congratulations for all your perfect job!!
Can you explain how fast is your new server?
Processor,ram,etc??
Run some benchmarks??
Thanks!
:)
Roland September 18, 2007 at 5:23 a.m.
The above error appear when I click "Timeline".
Roland September 18, 2007 at 5:23 a.m.
Trac Timeline seems to work now.
Xavi September 18, 2007 at 5:25 a.m.
Where do you have your old hosting solution?
Now, you have mediatemple: what solution: dv, dpv, cv, ...?
Thanks
Justin Bronn September 18, 2007 at 7:11 a.m.
I get a "500 Internal Server Error" when I try to commit... otherwise, it seems pretty snappy.
iki September 18, 2007 at 9:23 a.m.
Also http://www.djangoproject.com/document... seems not to work. Good news otherwise!
Brian September 18, 2007 at 12:32 p.m.
Very fast page loads from the East Coast. Fantastic job.
David Sissitka September 18, 2007 at 1:43 p.m.
Definitely an improvement. :) Somehow I was expecting Joyent/TextDrive though...
Jacob September 18, 2007 at 4:11 p.m.
Thanks for the bug reports, folks; they should have been fixed now.
Xavi: we've got something similar to the "dpv Nitro", though (mt) hooked us up with a custom box because we're picky about our hardware.
Mike September 18, 2007 at 4:33 p.m.
Congrats on the move. Any reason why you didn't go with WebFaction ? From what I see on the mailing list I thought they were the experts at hosting Django ?
Jeff Wheeler September 18, 2007 at 5:31 p.m.
David: I expected Joyent too. I suppose they couldn't pass up free hosting, though. :)
Android September 18, 2007 at 8:16 p.m.
I guess you will be one of the few to have actually have any decent form of support from Media Temple.
Moses Ting September 18, 2007 at 10:40 p.m.
This is fantastic news. As Malcolm Gladwell would say, it looks like Django is close to reaching the tipping point as more and more people hear and use it.
Observer September 19, 2007 at 1:45 a.m.
"I guess you will ..." Ah, I love the smell of bitter complaints from anonymous cowards in the morning.
Pete September 19, 2007 at 2:49 a.m.
Android: I agree. From what I've seen it seems like MediaTemple has a policy of giving special treatment to VIPs (Django, TechCrunch, ...) while the rest of us get slow and unhelpful support :(
Paul Bx September 19, 2007 at 9:49 a.m.
Very cool! I'll hold off a couple days on filing tickets about busted Trac macros :)
Alex from (mt) September 19, 2007 at 1:50 p.m.
Android/Pete Bx: If you are (mt) customers and have specific ticket numbers and complaints then email me ASAP at alex@mediatemple.net
If you aren't (mt) customers I'd like to hear from you and what compelled you to make the generalizations "the rest of us" and "one of the few".
Olive September 20, 2007 at 1:38 a.m.
Is Community Blog Aggregator still working ?
There is no new entry since September 17.
Andre Foulem September 20, 2007 at 9:39 a.m.
@Alex from (mt): A one click Django installer similar to what Webfaction have, would be great for the upcoming 'Django on the Grid(mt)'.
Zero September 20, 2007 at 10:24 a.m.
Olive: I was wondering the same thing. Also, where's the big announcement about The Django Book? For the love of Mike, it's available now!!
Walker Hamilton September 20, 2007 at 3:58 p.m.
I've got an (mt) GS account and am loving it! Can't wait for the Django containers.
Also, I help run Digital Web. I must say, (mt)'s support for DW & my personal acct has been helpful, fast, and unwavering.
Go (mt)!
Pinguino September 21, 2007 at 6:26 a.m.
@ Alex from (mt): WebFaction is definitely the example to follow in Django hosting.
@ Zero: The DjangoBook available? Where? On the Apress/Springer website I read "Available: November 28, 2007". I think it will be a perfect Christmas gift for the community. However don't be too anxious about the paper book: the online doc is already super.
Aktuariusz September 21, 2007 at 6:56 a.m.
Are there any converterters ne can you to transform existing site in php to django? I would like to go to django but there is a lot things to learn in python. And I would like to have easy start.
http://halley.pl
Warrant September 21, 2007 at 7:54 a.m.
No there aren't. Maybe you should start with drupal if you don't know python
Zero September 21, 2007 at 10:13 a.m.
Pinguino: My bad. Apress.com wasn't available yesterday morning, and Amazon.com was selling it and saying it was in stock and it could ship in two days. I ordered ... and today it says it will ship between October 23, 2007 - November 6, 2007. I'm betting that moves to November 28 :(
Alex (mt) September 21, 2007 at 2:37 p.m.
@Andre: Announcement is coming very soon on our blog. ;-)
@Pinguino: Webfaction does some great stuff. Agree 100%. I think people will like our setup/environment when we announce it...my hope is our future beta team gets granular and helps build a great Django container.
Alex (mt) September 21, 2007 at 5:14 p.m.
@Andre and anyone else who is interested, http://tinyurl.com/2bw2hu
we have more info and we're accepting beta applications from Grid users.
Pete September 24, 2007 at 4:04 a.m.
I've said it many times and I'll say it again: MediaTemple's "grid" is not a grid !!!
If you think your Django app will use the power of multiple machines on their system think again !
A real "grid" would require a much more advanced technology that only a few people have, and MediaTemple is certainly not one of them !
I'm not saying MediaTemple's system is bad, but if only they stopped their marketing lies they would gain much more respect from me.
Tom Dorr from ASO explains it better than me at http://www.hostdisciple.com/2006/11/2... :
"I hate marketing terms with the burning passion of a thousand suns.
Let me make this very clear: MT’s “Grid Server” is not a grid. It’s never been a grid. It never will be. It’s a cluster. It’s very simple to differentiate between the two types. In a grid, your processing task is parallelized between all the nodes in the grid. In a cluster, the processing task is done on one machine at a time, with the ability to failover to another machine, if needed. It’s horizontal versus vertical, basically. Grid computing is used all the time in scientific tasks and things like 3d rendering. Clusters are used when you cannot take a task and split it up into chunks to be processed on multiple nodes. When you request a page on MT’s “grid”, it’s just pulling from storage, through the webserver, out to the load balancer, and then is on it’s way to your computer. You cannot make this a parallel task, so it’s not a grid. There’s no way, nor any reason, to split out the serving of a single webserver request among multiple machines."
Pete.
Observer September 24, 2007 at 6:03 a.m.
"Pete", please take your trolling elsewhere. If you have a grind to axe, get a blog.
Xavi September 24, 2007 at 9:56 a.m.
Jacob, thanks for your info.
Nip September 24, 2007 at 10:36 a.m.
"Observer", I would not define Pete's statements as trolling. He makes a valid point, be it very opinionated. Besides, he marks his opinions clearly as such.
Pete September 25, 2007 at 3:32 a.m.
Observer: sorry if that sounded like a Troll, but I think I'm just stating the facts. By saying "Hundreds of servers for the price of one", they lead you to believe that your Django app will somehow use the power of multiple servers, which is simply not the case.
You only get your allocated cpu and memory usage on a single server and if your machine crashes there is a failover to another machine.
Chuck September 25, 2007 at 5:18 a.m.
After using MT for a year, I've since said goodbye.. I see it as some stupid little fashion label. Just hype, no value. Get those two little letters on your site "MT", and you are apparently very cool!
Alex (mt) September 25, 2007 at 6:21 p.m.
@Pete: You're really invested in making false claims about (mt), our GridContainer and Grid-Service messaging.
The Django GridContainers will have a detailed page explaining what it is. Our MySQL GridContainer page is a perfect product example: http://www.mediatemple.net/webhosting...
Our Django labs/beta sign-up page states it clearly in the 2nd bullet and in the intro paragraph: http://www.mediatemple.net/labs/grid/...
It's also important to note that when using a GridContainer for Ruby, Django, or MySQL, your site will still take advantage of the clustered web, mail, and ftp services deployed on the Grid platform; your application-specific tasks (e.g. Django) process in an isolated environment with dedicated resources, and then is served out by the Grid in the same way as any other site. The only way to provide this sort of flexibility in technologies in a practical manner is by using containers. (mt) Media Temple sees the potential in this technology and has invested time, money, blood, sweat, and tears into making a superior platform to host with.
If that's not enough and someone feels mislead, we implore you to take advantage of our money back guarantee and find a solution that you're comfortable with.
As for the Grid name, and Mr. Dorr's best guess last year, we, after internal debates, named it for the general public. Grid-Service is easier for customers to understand than "a cluster of clusters". As an added bonus, the other geographic locations we plan to add will push it closer to what some use to define a grid...although the grid debate continues - http://www.gridtoday.com/02/1209/1008...
Andre September 25, 2007 at 7:54 p.m.
Im really looking forward to experiment with GeoDjango.
oversize September 28, 2007 at 4:19 a.m.
Hello,
im happy for django to hear that.
This may be offtopic but something urged me to drop this note somewhere: "Django Rocks!!"
Thank You
Fredrik S September 28, 2007 at 2:40 p.m.
Please, give us some information about Django 1.0 and the book! :)
twiddlin thumbs September 30, 2007 at 10:23 a.m.
Amazon.com has the publication date for the Apress book pushed back to Nov 26 (I just checked); however they also listed it as "usually delivered in 2-3 days" for a brief period in mid September, so I'd take that new date with a grain of salt.
The videos at http://showmedo.com/videos/Django might help to tide you over (the 20 min wiki is wonderfully informative).
Adam S September 30, 2007 at 11:34 a.m.
I don't see how the "grid" vs. "cluster" issue is such a big thing. Really, I want to know more important things. For example, if I have a 64MB container, how many boxes have that container. If 4 boxes have it, it's going to perform really well when you load balance between them. If it's 2 boxes, I'm not really getting much of a clustering effect.
Really, what are the advantages of this cluster setup? Redundant hardware (which minimizes downtime) seems to be the big seller to me. But what about performance. mt says that resource limits are history, but that just isn't true - there are always limits, this just might have better limits. Heck, the RAM limit is pretty specific and RAM is probably the most important resource on a server.
So mt seems to be withholding the most important piece of information when it comes to judging how well their service would work. If I have a 64MB container, how many boxes is it running on (behind that load balancer)? If it's running on 8 boxes, I pretty much have the equivalent of a 512MB VPS. If it's running on 2 boxes, it's a very different story.
Or maybe the cluster creates and destroys containers based on site load ans so they can't give us this information? Then the question becomes, how well does mt react to bursts. Does it take 10 minutes of sustained load to get another container created? Does it happen anytime the system sees response times over 500ms?
Even something as simple as their GPU measurement. What is a GPU? Is one GPU equivalent to about 1 minute of CPU time on a 1GHz processor? Is this just some measurement with no basis that mt can say I'm going over when they feel like more money?
I'm not saying whether mt's service is good or bad. What I am trying to point out is that, as a prospective customer, I have no way of judging what mt's service is. With a VPS, I can pretty easily tell what I'm getting. With mt, I could be getting anywhere from 64MB on 1-2 machines to 64MB on 1000 machines. When you're a customer, do you get to see in the control panel how many boxes contain your container at a given time?
It's just very black box - a "trust us, you won't suffer the slashdot effect. how do we know this? well, um, it's grid based and utility and replicated and clustered. those words should placate you."
I'm not trying to flame mt. In fact, this should give mt an idea of what they need to tell prospective customers who are a little above those who are happy with buzzwords. If I had more information, I'd probably become an mt customer, but I can't just sign up for something with no indication of its true specs.
Alex (mt) October 2, 2007 at 6:45 p.m.
@Adam S: Regarding clustering, the (gc) Grid-Containers themselves are not clustered. However they do sit behind the (gs) Grid-Service which allows you to take advantage of clustering for all the applications you have outside of your Container. The concept behind Containers is to offer specialized server environments for applications such as Django, RoR and MySQL in a way that strikes a balance between a fully customized, "from the ground" setup you would have to put together yourself (standard VPS server) or a more limited, restrictive environment like the ones you find from traditional shared hosting providers. Containers provide guaranteed resources along with most of the flexibility of a VPS server without dealing with the initial setup or the on-going maintenance it usually takes run these specialized application environments.
To answer the GPU question (from this Kb article: http://kb.mediatemple.net/article.php...) 1 GPU = 7.24% load on one processor for 1 hour. Based on a 30 day cycle, 1000 GPU's allows you to run 1.38 GPU's per hour. And GridContainer processing is not counted for GPU usage, so if you have heavy Django processes in your container, it does not count towards GPU usage.
I hope I've been able to provide some clarity.
D.A.Ridgeway October 5, 2007 at 11:55 a.m.
>Bonus points to anyone who gets the reference for the title...
From the remake of Lady Killers
The woman who owns the house that Tom Hanks character rents.
She complains about the rap song, "I left my Wallet in El Segundo"
So what do I get? A free copy of Django (wait a minute!?!?!)
Comments are closed
To prevent spam, comments are no longer allowed after sixty days.



Matthias Pronk September 17, 2007 at 6:46 p.m.
Reference is probably to this song:
http://en.wikipedia.org/wiki/I_Left_M...
:)