Slashdot Mirror
Use apt-p2p To Improve Ubuntu 9.04 Upgrade
An anonymous reader writes "With Jaunty Jackalope scheduled for release in 12 days on April 23, this blog posting describes how to switch to apt-p2p in preparation for the upgrade. This should help significantly to reduce the load on the mirrors, smooth out the upgrade experience for all involved, and bypass the numerous problems that have occurred in the past on Ubuntu release day. Remember to disable all third-party repositories beforehand."
The site doesn't have much information, but other sources I have read state that apt-p2p is very experimental. Use at your own peril!
Put identity in the browser.
You can also upgrade Ubuntu with an alternate install CD. These can be downloaded via bittorrent, and usually trigger an "automatic update" prompt as soon as they are inserted into an existing Ubuntu system.
I had wondered for a while why yum and apt did not do this by default. It would seem a great ideal. However.... I recently tried to down load fedora 11 alpha via bit torrent using a BT internet connection in the UK. It worked great for about 10Mb (@90-100kb/s), then the download speed gradually ground to a halt. (5kb/s) When I tried a direct download of the same iso the speed bumped back up to a steady 100kb/s. I concluded BT was throttling my bit torrent connection of a legal download to a very slow speed.
So my point is sounds like a great idea but if it is enabled by default it had better have some way to detect bandwidth throttling of p2p networks and revert to http transfer.
mirror here: http://74.125.77.132/search?q=cache:3gY3Bq4EKnMJ:blog.chenhow.net/os/linux/ubuntu/using-apt-p2p-for-faster-upgrades-from-intrepid-to-jaunty/+http://blog.chenhow.net/os/linux/ubuntu/using-apt-p2p-for-faster-upgrades-from-intrepid-to-jaunty&cd=1&hl=nl&ct=clnk&gl=nl
I have yet to have an Ubuntu distro update smoothly, ever. But that won't stop me, onward I will plunge headlong into it with abandon. I don't like my data anyway.
What I like about this is not so much the potentially faster upgrade as the ability to contribute a bit to others. The six-monthly upgrades are are rate enough that I don't mind if they are a bit slow - not that they have been. But I am very conscious that I am using other people's freely given bandwidth and I am pleased to be able to give some back.
Does anybody know if I can force my various machines to cross-peer from each other? If I update one first, I don't want the others searching the Net for peers - they should just copy from the first.
Consciousness is an illusion caused by an excess of self consciousness.
I'm concerned that after reading the article, and apt-p2p's FAQ page, that I can't find any guide to how much upload bandwidth this thing will use. While I'm all for sharing, I find it important to cap my upload speed so my connection performs well on other stuff I'm doing, and also stop uploading once I'm at 1:1 sharing or so. Some of us pay if we use too much bandwidth!
Combination - fun iPhone puzzling
p2p is a method used exclusively by criminals, there's no way im going to be using this method.
It worked for me. But in case it really is slashdotted here's the story, from memory (let's test those theories eh?)
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Many primary Linux download sites wind up taking an unreasonable amount of traffic from default setups. If you want to contribute back to the OS's and packages that you find so useful, consider setting up a local mirror to share with the world at large. If you can't justify that, at least consider setting up an internal rsync mirror anytime you have a dozen or more boxes to make updates and downloads much faster for your site, and configure your local machines to point to that local mirror.
This turns out to be especially useful for PXE installaters and cluster setups, for any Linux or other OS. There's nothing like having 100 internal Linux machines all trying to update OpenOffice at the same time from an external primary site, through a corporate DSL line, to ensure that many of the updates will fail.
If it works, why upgrade at all?
Ubuntu 8.04 is a Long Term Support (LTS) release. It will have any security patches until the next LTS release, which is typically every 18 months. So, why not just wait for 9.10?
ws
So does Anonymous Coward have good karma?
that a site advising the use of p2p to prevent the meltdown of servers has itself been slashdotted.
On a side note : web data and pages themselves could be p2p distributed too, no? Say a peer gets a webpage's hash (containing html and images) and the date/time of expiry for a webpage from a server. If other peers have that page (html+images), and it's up to date, you could download their copy. Otherwise, the server sends a fresh copy to you, and you seed it for others. Not being in computer science, I'm sure this has been proposed before and that there are glaring shortcomings I have missed.
Me too. Often it's quicker to do a full install from scratch.
That's why my systems always have at least two different partitions: one for "/" and another for "/home". I can reformat my system partition and still have my data intact.
"Is it just me or is the fun game of "pick your closest mirror" not very fun at all? Just download the damn thing at best possible speed. I don't care where you get it from. "
You are aware that "closest" in this context means "faster", aren't you?
"As if I'm in a position to pick the best site where to download something from. Give me a break. Apologies to the power users who can lick their Ethernet cable and tell which site will have the best download performance and availability."
Probably is too much a power user the one able to install the package "apt-spy" which will build a sources.list for you based on bandwith probes, isn't it?
Oh, and please, don't let parent post at +Insigthful when it's plain -Nonsense.
It's not funny and it's not sarcasm. It's the same cliche meme repeated a thousandth time. LOL, p2p is helping terrorism, mod me funny. I could write a python one-liner that would produce more varied content than most of these +5, funnies all over here on /.
I even actually wrote it:
$ python -c 'print (lambda words, random: " ".join([random.choice(words)[:-1] for i in range(random.randint(1,10))]))(open("/usr/share/dict/words").readlines(), __import__("random"))+"."'
angoras lawgiver's Father's approbations uninteresting inferring Antonio's Clotho's chlorine.
Is your point that a host that's connected via T-1 that's a mile away is faster than a host that's connected on an OC-3 3,000 miles away? That is, based on knowing the geographic location of a host, you're saying it's somehow an indication of how fast my download will complete? That's the only thing that matters to me -- when will I have my completed bits. My only point here is that the information given in mirror selection is not enough to pick the "fastest" way to get what I want. It lists the geographical location and that's it.
Yes. Yes it is. If such a list can be generated, then why not just generate it in normal operation or list the mirrors based on the output of that tool? Though I do appreciate the tip, and I will try it. Obviously bandwidth and availability varies on a day by day basis. So taking a snapshot at one point in time seems like it will get stale.
It's a legitimate end-user concern. "Which mirror should I select" should not be a user problem. The user wants his bits as soon as possible, which is a technical problem that has allegedly been solved with apt-spy. If that's the case, we should probably integrate that with the mirror selection process, and then you don't have to put up with all the "Nonsense".