271

Unfortunately, our hosting provider experienced 100% data loss, so I've lost all content for two hosted blog websites:

(Yes, yes, I absolutely should have done complete offsite backups. Unfortunately, all my backups were on the server itself. So save the lecture; you're 100% absolutely right, but that doesn't help me at the moment. Let's stay focused on the question here!)

I am beginning the slow, painful process of recovering the website from web crawler caches.

There are a few automated tools for recovering a website from internet web spider (Yahoo, Bing, Google, etc.) caches, like Warrick, but I had some bad results using this:

  • My IP address was quickly banned from Google for using it
  • I get lots of 500 and 503 errors and "waiting 5 minutes…"
  • Ultimately, I can recover the text content faster by hand

I've had much better luck by using a list of all blog posts, clicking through to the Google cache and saving each individual file as HTML. While there are a lot of blog posts, there aren't that many, and I figure I deserve some self-flagellation for not having a better backup strategy. Anyway, the important thing is that I've had good luck getting the blog post text this way, and I am definitely able to get the text of the web pages out of the Internet caches. Based on what I've done so far, I am confident I can recover all the lost blog post text and comments.

However, the images that go with each blog post are proving…more difficult.

Any general tips for recovering website pages from Internet caches, and in particular, places to recover archived images from website pages?

(And, again, please, no backup lectures. You're totally, completely, utterly right! But being right isn't solving my immediate problem… Unless you have a time machine…)

29
  • 101
    When somebody like Jeff Atwood himself can lose two entire websites in one fell swoop... Well. I'm going to review my own backup procedures, for one :P
    – Phoshi
    Dec 11, 2009 at 21:10
  • 243
    @Phoshi: Jeff has some good articles on Coding Horror on backup. You should give them a quick read.
    – joshhunt
    Dec 11, 2009 at 21:31
  • 36
    joshhunt wins one (1) internet. This offer may not be combined with other offers, exchanged, or substituted. No rainchecks.
    – Adam Davis
    Dec 11, 2009 at 21:50
  • 30
    The lengths some people will go to, to earn rep on SU...
    – Antony
    Dec 12, 2009 at 0:11
  • 28
    Please don't refer to what you did as "backups" - if those files are on the same server, they're in no way "backups."
    – wahnfrieden
    Dec 14, 2009 at 22:35

42 Answers 42

225

Here's my wild stab in the dark: configure your web server to return 304 for every image request, then crowd-source the recovery by posting a list of URLs somewhere and asking on the podcast for all your readers to load each URL and harvest any images that load from their local caches. (This can only work after you restore the HTML pages themselves, complete with the <img ...> tags, which your question seems to imply that you will be able to do.)

This is basically a fancy way of saying, "get it from your readers' web browser caches." You have many readers and podcast listeners, so you can effectively mobilize a large number of people who are likely to have viewed your web site recently. But manually finding and extracting images from various web browsers' caches is difficult, and the entire approach works best if it's easy enough that many people will try it and be successful. Thus the 304 approach. All it requires of readers is that they click on a series of links and drag off any images that do load in their web browser (or right-click and save-as, etc.) and then email them to you or upload them to a central location you set up, or whatever. The main drawback of this approach is that web browser caches don't go back that far in time. But it only takes one reader who happened to load a post from 2006 in the past few days to rescue even a very old image. With a big enough audience, anything is possible.

6
  • 53
    +1 for the most creative approach. Could actually work since CH has some many readers.
    – delux247
    Dec 11, 2009 at 22:17
  • 17
    implemented here? diovo.com/2009/12/… Dec 14, 2009 at 21:18
  • 3
    I think you could crawl your static files for the image tags and copy all of those into one giant page of images, instead of having everybody click each link. The diovo.com implementation looks very impressive, hope it works out for you.
    – phloopy
    Dec 15, 2009 at 6:00
  • OMG! Very nice analysis. Aug 7, 2011 at 14:41
  • 4
    In fact, you should be able to retrieve images using canvas and send them home by AJAX. May 20, 2014 at 17:24
67

Some of us follow you with an RSS reader and don't clear caches. I have blog posts that appear to go back to 2006. No images, from what I can see, but might be better than what you're doing now.

7
  • +1 definitely. Google Reader doesn't, but I bet a desktop-based one would.
    – Nicolas Webb
    Dec 11, 2009 at 21:01
  • 3
    You could also ask people to check their browser caches. Those who view Coding Horror retro-style might have some of the images cached.
    – Alex Rozanski
    Dec 11, 2009 at 21:04
  • I've got blog posts back to 2005 in GReader, but unfortunately, they don't have images, and they won't let me just export those as a series of pages... I could email them to you though, Jeff... Dec 11, 2009 at 21:08
  • Yeah, there was an implied "I'll send you what I have if you ask for it." in my answer as well.
    – retracile
    Dec 11, 2009 at 21:09
  • 4
    Too many RSS readers assume images will never die. I know mine does :(
    – Phoshi
    Dec 11, 2009 at 21:14
63

(1) Extract a list of the filenames of all missing images from the HTML backups. You'll be left with something like:

  • stay-puft-marshmallow-man.jpg
  • internet-properties-dialog.png
  • yahoo-homepage-small.png
  • password-show-animated.gif
  • tivo2.jpg
  • michael-abrash-graphics-program

(2) Do a Google Image Search for those filenames. It seems like MANY of them have been, um, "mirrored" by other bloggers and are ripe for the taking because they have the same filename.

(3) You could do this in an automated fashion if it proves successful for, say, 10+ images.

1
  • 2
    Would be very ironic if he actually got any images back like this. Sep 6, 2019 at 22:47
51

By going to Google Image search and typing site:codinghorror.com you can at least find the thumbnailed versions of all of your images. No, it doesn't necessarily help, but it gives you a starting point for retrieving those thousands of images.

Codinghorror images

It looks like Google stores a larger thumbnail in some cases:

Google vs. Bing

Google is on the left, Bing on the right.

5
  • 2
    yeah, worst case, we'll have to scale up the thumbnails from Google. I hear Bing stores larger thumbnails, though? Dec 11, 2009 at 20:59
  • I don't know; I'm not a bing sort of guy. I don't even know if they do Image search like Google does. I'll find out and update said post. Dec 11, 2009 at 21:03
  • 19
    I don't know if this is you. But Imageshack seems to have many of your blog images. profile.imageshack.us/user/codinghorror Dec 11, 2009 at 21:04
  • They seem to have what looks like 456 images that are full size. This might be the best bet for recovering everything. Maybe they can even provide you a dump. Dec 11, 2009 at 21:09
  • 29
    Use the Google thumbnails as a start, then use tineye.com to see if anyone is hosting a copy.
    – sep332
    Dec 11, 2009 at 21:48
41

Sorry to hear about the blogs. Not going to lecture. But I did find what appears to be your images on Imageshack. Are they really yours or has somebody been keeping a copy of them around.

http://profile.imageshack.us/user/codinghorror

They seem to have what looks like 456 images that are full size. This might be the best bet for recovering everything. Maybe they can even provide you a dump.

39

Jeff, I have written something for you here

In short what I propose you do is:

  1. Configure the web server to return 304 for every image request. 304 means that the file is not modified and this means that the browser will fetch the file from its cache if it is present there. (credit: this SuperUser answer)

  2. In every page in the website, add a small script to capture the image data and send it to the server.

  3. Save the image data in the server.

  4. Voila!

You can get the scripts from the given link.

0
29

Try this query on the Wayback Machine:

http://web.archive.org/web/*sa_re_im_/http://codinghorror.com/*

This will get you all the images from codinghorror.com archived by archive.org. This returns 3878 images, some of which are duplicates. It will not be complete, but a good start none the less.

For the remaining images, you can use the thumbnails from a search engine cache, and then do a reverse look-up using these at http://www.tineye.com/ . You give it the thumbnail image, and it will give you a preview and a pointer to closely matching images found on the web.

2
27

+1 on the dd recommendation if (1) the raw disk is available somewhere; and (2) the images were simple files. Then you can use a forensic 'data-carving' tool to (for example) pull out all credible ranges that appear to be JPGs/PNGs/GIFs. I've recovered 95%+ of the photos on an iPhone that was wiped this way.

The open source tools 'foremost' and its successor 'scalpel' can be used for this:

http://foremost.sourceforge.net/

http://www.digitalforensicssolutions.com/Scalpel/

2
  • 3
    Photorec may also be of use once you get dd images.
    – Broam
    Dec 11, 2009 at 22:28
  • foremost is available via yum on Fedora
    – retracile
    Dec 12, 2009 at 16:50
27

Luckily, future generations will be ok.

Even with only some of this big rock, scientists/linguiststs figured out a lot.

Rosetta Stone

If a few pictures are missing, leave it to someone to figure out in a couple thousand years.

Hopefully, you're laughing a little. :)

1
  • 6
    Ok, you got a giggle from me at least ;-)
    – Ogre Psalm33
    Dec 16, 2009 at 17:09
22

You could always try archive.org, as well. Use the wayback machine. I've used this to recover images from my websites.

4
  • 3
    Doesn't seem to have much of a cache for CodingHorror, at least. I do see images for blog.stackoverflow though.
    – The How-To Geek
    Dec 11, 2009 at 21:00
  • i rebuilt a website using internet wayback machine once but i tried a few times since and it really doesn't archive very many sites...
    – djangofan
    Dec 11, 2009 at 21:35
  • Looks like it goes back to 2004 here web.archive.org/web*/codinghorror.com
    – Chris Nava
    Dec 11, 2009 at 23:41
  • Thank goodness it didn’t have a robots.txt file huh? :)
    – Synetech
    Dec 12, 2009 at 20:06
15

So, absolute worst case, you can't recover a thing. Damn.

Try grabbing the minified google ones, and putting them through TinEye, the reverse-image search engine. Hopefully it should grab any duplicates or rehosts people have made.

15

It is a long shot, but you could consider:

  • Posting the exact list of picture you are missing
  • crowd-sourcing the retrieval process through all your readers's internet cache.

For instance, see the Nirsoft Mozilla Cache Viewer:

alt text
(source: nirsoft.net)

It can quickly dig up any "blog.stackoverflow.com" picture one might still have through a simple command line:

MozillaCacheView.exe -folder "C:\Documents and Settings\Administrator\Local Settings\Application Data\Mozilla\Firefox\Profiles\acf2c3u2.default\Cache" 
/copycache "http://blog.stackoverflow.com" "image" /CopyFilesFolder "c:\temp\blogso" /UseWebSiteDirStructure 0

Note: they have the same cache explorer for Chrome.

alt text
(source: nirsoft.net)

(I must have 15 days worth of blog.stackoverflow.com pictures in it)

And Internet Explorer, or Opera.


Then update the public list to reflect what the readers report finding in their cache.

13

In the past I've used http://www.archive.org/ to pull up cached images. It's kind of hit or miss but it has worked for me.
Also, when trying to recover stock photos that I've used on an old site, www.tineye.com is great when I only have the thumbnails and I need the full size images.

I hope this helps you. Good Luck.

2
  • I looked through archive.org a few minutes ago for codinghorror.com images and the few posts I clicked didn't have any showing. Dec 11, 2009 at 20:58
  • Archive.org releases the data months after they first indexed them.
    – Christian
    Dec 12, 2009 at 15:39
11

This is probably not the easiest or most full-proof solution, but services like Evernote typically save both the text and images when they are stored inside the application - maybe some helpful readers who saved your articles could save the images and send them back to you?

11

I've had great experiences with archive.org. Even if you aren't able to extract all of your blog posts from the site, they keep periodical snapshots:

alt text

This way you can check out each page and see the blog posts you made. With the names of all the posts you can easily find them in Google's cache if archive.org doesn't have it. Archive tries to keep images, Google cache will have images, and I haven't emptied my cache recently so I can help you with the more recent blog posts :)

2
  • I tried to get some data from the website of a company I used to work for a while ago. It was good for the text, less so for the images. But YMMV
    – ChrisF
    Dec 11, 2009 at 22:58
  • I believe Google web cache does not store images.
    – Nathaniel
    Dec 13, 2009 at 0:51
8

A suggestion for the future: I use Windows Live Writer for blogging and it saves local copies of posts on my machine, in addition to publishing them out to the blog.

1
  • Plus, using Windows Live Writer is just good common sense.
    – The How-To Geek
    Dec 12, 2009 at 3:33
8

Have you tried your own local browser cache? Pretty good chance some of the more recent stuff is still there. https://lifehacker.com/resurrect-images-from-my-web-browser-cache-33300382

(Or you could compile a list of all missing images and everyone could check their cache to see if we can fill in the blanks)

7

About five years ago, an early incarnation of an external hard drive on which I was storing all my digital photos failed badly. I made an image of the hard drive using dd and wrote a rudimentary tool to recover anything that looked like a JPEG image. Got most of my photos out of that.

So, the question is, can you get a copy of the virtual machine disk image which held the images?

7

The web archive caches the images. It's under heavy load right now, you should be ok until 2008 or so.

http://web.archive.org/web/20080618014552rn%5F2/www.codinghorror.com/blog/

7

I suggest the combination of archive.org and a request anonymizer like [Tor][2]. I suggest using anonymizer because that way each of your requests will have a random IP and location and that way you can avoid getting banned by a archive.org (like Google did) for unusually high number of requests.

Good Luck, there are a lot of gems in that blog.

1
  • Given that Jeff wants to make a donation to archive.org, so abusing the anonymizer might not be absolutely innacceptable. But I still want give you a kick for that. :-|
    – hstoerr
    Dec 15, 2009 at 7:14
6

The wayback machine will have some. Google cache and similar caches will have some.

One of the most effective things you'll be able to do is to email the original posters, asking for help.

I do actually have some infrastructural recommendations, for after this is all cleaned up. The fundamental problem isn't actually backups, it's lack of site replication and lack of auditing. If you email me at the private email field's contents, later, when you're sort of back on your feet, I'd love to discuss the matter with you.

6

If your images were stored on an external service such as Flickr or a CDN (as mentioned in one of your podcasts), you may still have the image resources there.

Some of the images could be found searching on Google Images and click on "Find similar images", maybe there are copies on other sites.

5

archive.org sometimes hides images. Get each URL manually (or write a short script) and query them for it like this:

string.Format("GET /*/{0}", nextUri)

Of course that's going to be quite a pain to search through.

I might have some in my browser cache. If I do I'll host them somewhere.

4

If you're hoping to try to scrape users' caches, you may want to set the server to respond 304 Not Modified to all conditional-GET ('If-Modified-Since' or 'If-None-Match') requests, which browsers use to revalidate their cached material.

If your initial caching headers on static content like images were pretty liberal -- allowing things to be cached for days or months -- you could keep getting revalidate requests for a while. Set a cookie on those requests, and appeal to those users to run a script against their cache to extract the images they still have.

Beware, though: the moment you start putting up any textual content with inline resources that aren't yet present, you could be wiping out those cached versions as revalidators hit 404s.

4

You could use TinEye to find duplicates of your images by searching the thumbnails with google cache. This will help only with images you've taken from others site, though.

2
  • 1
    No, it would help with images others have taken from CH. Dec 24, 2009 at 23:24
  • @DisgruntledGoat: I didn't even thought of that at first :D
    – voyager
    Jan 6, 2010 at 0:10
4

At the risk of pointing out the obvious, try mining your own computer's backups for the images. I know my backup strategy is haphazard enough that I have multiple copies of a lot of files hanging around on external drives, burned discs, and in zip/tar files. Good luck!

4

I've managed to recover these files from my Safari cache on Snow Leopard:

bad-code-offset-back.jpg
bad-code-offset-front.jpg
code-whitespace-invisible.png
code-whitespace-visible.png
coding-horror-official-logo-small.png
coding-horror-text.png
codinghorror-search-logo1.png
crucial-ssd-128gb-ct128m225.jpg
google-microformat-results-forum.png
google-microformat-results-review.png
kraken-cthulhu.jpg
mail.png
powered-by-crystaltech-web-hosting.png
ssd-vs-magnetic-graph.png

If anyone else wants to try, I've written a Python script to extract them to ~/codinghorror/filename, which I've put online here.

I hope this helps.

3

Did you get a chance to see if, your hosting provider has any backup at all (some older versions)?

1
  • it does not look good.. their backup program was unable to backup the virtual machine hard drive files, so there are no backups. Dec 11, 2009 at 21:03
2

How much is this data worth to you? If it's worth a significant sum (thousands of dollars) then consider asking your hosting provider for the hard drive used to store the data for your website (in the case of data loss due to hardware failure). You can then take the drive to ontrack or some other data recovery service to see what you can get off the drive. This might be tricky to negotiate due to the possibility of other people's unrecovered data on the drive as well, but if you really care about it you can probably work it out.

3
  • the server was a VM as far as I know.
    – splattne
    Dec 11, 2009 at 21:32
  • 1
    @splattne even so, there's a non-zero chance a lot of the data could be recovered.
    – Wedge
    Dec 11, 2009 at 22:05
  • Would have to be a highly specialised service.
    – Phoshi
    Dec 11, 2009 at 22:10
2

Very sorry to hear this and I am very annoyed for you, and the timing - I wanted an offline copy of a few of your posts and did HTTrack on your entire site but had to go out (this was a couple of weeks ago) and I stopped it.

If the host is half descent - and by the fact I am guessing you are a good customer... I would ask them to either send you the hard drives (as I am guessing they should be using RAID) or do some recovery themselves.

Whilst this may not be a fast process, I did this with one host for a client and was able to recover entire databases intact (... basically, the host tried an upgrade for the control panel they were using and messed it up.. but nothing was overwritten).

Whatever happens - Good luck from all your fans on the SO sites!

Not the answer you're looking for? Browse other questions tagged or ask your own question.