SSH public RSA key errors

Seen these before when trying to login via SSH with your new RSA public key?

Nov 2 12:09:17 hostname sshd[12712]: error: buffer_get_ret: trying to get more bytes 257 than in buffer 73
Nov 2 12:09:17 hostname sshd[12712]: error: buffer_get_string_ret: buffer_get failed
Nov 2 12:09:17 hostname sshd[12712]: error: buffer_get_bignum2_ret: invalid bignum
Nov 2 12:09:17 hostname sshd[12712]: error: key_from_blob: can't read rsa key
Nov 2 12:09:17 hostname sshd[12712]: error: key_read: key_from_blob AAAAB3N[...] failed

In my case these were the result of copying a public key from e-mail, which tends to mangle long text lines. I usually don’t have this problem because I use the ssh-copy-id script to copy my keys to a remote host before attempting to log in.

Apache custom logging

Aren’t you interested in seeing what requests users, bots, or script kiddies make of your site, especially those things that client-side JavaScript-based analytics packages don’t tell you?

Under Apache, custom logging can give you lots of information you might not have seen otherwise. I’ll let the documentation for Apache’s mod_log_config say most of this, but as a quick preview, you could try defining a custom log format up near the top of your httpd.conf with

LogFormat "%a %t %{Host}i \"%r\"" hostlog

for example, then in all of your Directory containers, you could do

CustomLog logs/forest-monsen-site-host-log hostlog

Then, in my case, /var/log/httpd/forest-monsen-site-host-log would contain lines like [31/Aug/2010:08:53:24 -0500] "GET /aggregator/sources/2 HTTP/1.0" [31/Aug/2010:08:53:24 -0500] "GET /images/house.gif HTTP/1.1"

And I’d be able to tell which hostname was originally requested by the user — before any of my mod_rewrite rules got to it. Good stuff.

sftp chroot jail in Ubuntu

(Update 16 Mar 2011: Since writing this post, I’ve learned of an easier way to create this chroot jail. Newer versions of OpenSSH enable the “ChrootDirectory” configuration directive. I recommend that you take a look at George Ornbo’s tutorial on chrooting sftp users in Intrepid for the details.)

(Updated 08 Feb 2011 to reflect xplicit’s experience on Ubuntu 10.04.)

I wanted to give a buddy access to a website hosted on my box. So I tried scponly, since I only wanted to provide SFTP access to that particular directory, using a chroot jail. The steps are as follows.

  1. Install the scponly package using Ubuntu’s APT package management system.
  2. Use the script provided to set up your first jail and your user’s home directory. For the location of the user’s jail, give the path of the directory you want to share.
  3. Provide a password for the new user.
  4. Ensure that the new user has permissions to read and write all the necessary directories in your Web site.

$ sudo apt-get install scponly
$ gzip -dc /usr/share/doc/scponly/setup_chroot/ > /tmp/
$ cp /usr/share/doc/scponly/setup_chroot/config.h /tmp

The previous step copies the “config.h” file to help things go more smoothly, as Luke found.

$ chmod +x /tmp/
$ cd /tmp
$ sudo ./

Next we need to set the home directory for this scponly user.
please note that the user's home directory MUST NOT be writeable
by the scponly user. this is important so that the scponly user
cannot subvert the .ssh configuration parameters.
For this reason, a writeable subdirectory will be created that
the scponly user can write into.

Note that I removed the /incoming subdirectory created by this script. There was no need for a separate directory for my buddy to upload files. He could have permissions over the whole site tree.

-en Username to install [scponly]
-en home directory you wish to set for this user [/home/bob]
-en name of the writeable subdirectory [incoming]

creating /var/www/sites/bobsite/htdocs/incoming directory for uploading files

Your platform (Linux) does not have a platform specific setup script.
This install script will attempt a best guess.
If you perform customizations, please consider sending me your changes.
Look to the templates in build_extras/arch.
- joe at sublimation dot org

please set the password for bob:
Enter new UNIX password:
Retype new UNIX password:
passwd: password updated successfully
if you experience a warning with winscp regarding groups, please install
the provided hacked out fake groups program into your chroot, like so:
cp groups /var/www/sites/bobsite/htdocs/bin/groups

This script added certain directories to the site root (/var/www/sites/bobsite/htdocs). Every other directory needed to be writable by Bob. So let’s add Bob to a special group, and allow that group write access on all the website’s files.

$ sudo adduser bob www-data

We can ignore /bin, /etc, /lib and other directories added to the chroot jail (the website filesystem):

$ sudo find . \! -user root -exec chgrp www-data \{\} \;
$ sudo find . \! -user root -exec chmod g+w \{\} \;

Good to go!

Server move complete

I migrated a bunch of stuff from a CentOS 4 server to Ubuntu 8.04 LTS over the last couple of days.

  • Five websites: One Moodle and one Drupal site backed by MySQL databases, and three static sites. SSL setup.
  • Added some software. How can I work without vim and slocate?
  • Security hardening, including a service review, permissions, firewall setup, administrative access through SSH, sudo config, and Postfix with spam filtering.
  • Nagios server monitoring config.

I checked my work logs and decided that I did pretty well, considering I got it all done in 10 hours 35 minutes.

Set Debian or Ubuntu server timezone

This one’s an easy one, from the tzselect (1) manpage:

sudo dpkg-reconfigure tzdata

Flush DNS cache in Ubuntu

Interested in flushing your Ubuntu DNS cache? Note: I’m running Jaunty Jackalope as of the date of this post.

Well, Ubuntu doesn’t cache DNS by default. Your cache rests within your router, or your assigned DNS servers. You could restart your router, if you have access to it. Or wait until the time-to-live has expired.

You can install a local resolver that will cache DNS addresses, if you like. It will speed up your Web access slightly, since your Web browser will check the local cache first. I imagine the time you save will be measured in milliseconds.

Do that with:

sudo apt-get update && sudo apt-get install nscd

And to clear your local cache, restart the service:

sudo /etc/init.d/nscd restart

Recursively find and list filesize and full path on the command line

Can’t beat the command line for flexibility and power in accomplishing system administration tasks. Here’s one way to recursively list the filesizes and full paths of files with a particular extension from the command line:

nice find . -name "*.swf" -type f -print0 | xargs -0r ls -skS | less

This is a succinct way to say:
“Show me all Flash files in the current directory hierarchy, descending to unlimited depth. Print the full filename on standard output followed by a null character. Send each filename in turn to the ‘ls’ command, which will look up each file’s size and print that in 1K blocks followed by the filename. (If there aren’t any results from the first command, don’t even run the ‘ls’ command, since that will just give us a list of all the files in the current directory.) Finally, send all that output to the ‘less’ command, which will allow me to page through and view it easily.”

EDIT: Added -r switch to xargs command to ensure we don’t see a list of all files, if the first ‘find’ command doesn’t find any. That sort of thing could be confusing.