Archive for the ‘linux/unix’ Category

This is a relatively simple trick, but one that I thought was rather clever.

I needed to SSH into a target server (172.16.4.1) that only allowed connections from a specific IP address (172.16.4.10). I knew that the system holding that specific IP address was down, yet I still needed to SSH into the target.

To further complicate matters, I did not have physical access to the target network at the time. But, I did have remote access (and root) to another Linux box on the network.

Hmmm…

So, all I needed to do was change the IP address of my one good box to the one expected by the target system, but if I changed the IP address on the main interface, then I would disconnect myself.

Enter virtual interfaces and routing.

I created a new virtual interface with ifconfig (assuming 172.16.4.10 is the IP we want to come from):

/sbin/ifconfig eth0:1 172.16.4.10

This does not, however, mean that my SSH connection will come from this address, so I needed to change the routing table (assuming 172.16.4.1 is what we want to connect to):

/sbin/route add -host 172.16.4.1 eth0:1

Now, I will effectively be coming from my new interface when I connect to the target server.

So, today I decided to set about the task of backing up my Gmail to my home server. I pay Google for 99% uptime, and really don’t expect their service to fail, but I have to account for more likely scenarios. For example, I may conceivably delete a message and find out I need it later; or, I may not want to keep a message on Gmail anymore but still need to retain it.

The players in this setup are a server running CentOS 5.2 and Gmail itself. My plan is to create a cron job that backs up the email at some fixed interval (daily) and saves it to a folder. At some point, I will back that up to DVD in case I or Gmail lose everything one day.

Gmail Backup runs on Windows and Linux. Gmail Backup does something that fetchmail or other solutions fail to provide: your Gmail labels are actually stored. This made it appear to be a better solution overall, so I decided to give it a whirl.

It is written in Python and requires Python to run on Linux (the Windows version is compiled). Unfortunately for me, the Python that comes with CentOS 5.2 is Python 2.4.3 and Gmail Backup requires Python 2.5.

I could just rip out the Python 2.4.3 libraries and replace them with Python 2.5 libs, but that would create problems for some of CentOS’s native utilties. So, I’ve installed Python 2.5 and wxpython in a different directory on my server and pointed Gmail Backup at the correct Python files.

According to the documentation, I just needed to run:

gmail-backup.sh backup my-backup-directory gmail-username gmail-password

The first attempt at running Gmail Backup met with failure:

Error: IMAP: [ALERT] Invalid credentials (Failure)

which I realized was because the shell was not escaping the ! in my password, so I changed the password.

Then I got:

ValueError: time data did not match format:  data=XXXXX  fmt=%Y%m%d

The XXXXX was the last half of my new password, which had a space in it. I consider this a bug in the implementation of Gmail Backup. I’ll contact the author with a fix later. So, I removed the space from my password.

That did it. Off I went. Gmail Backup nicely put my email into a directory. Each message is stored as a text file in directories indexed by year and date. All of the labels are stored in a text file.

Restoring to Gmail is as easy as:

gmail-backup.sh restore my-backup-directory gmail-username gmail-password

With a little scripting and a cron job, now I can automate Gmail Backup with a script, having it create periodic backups of all of my messages, then compressing the backup directory into one big file for storage and offloading.

Thanks Gmail Backup for making this so much easier!

So, as noted in earlier posts, I decided the best way to share files between my Ubuntu 8.10 and Windows XP installations on my laptop was to use an NTFS partition for my home directory/My Documents.

Unfortunately, applications like wine issue serious warning messages or refuse to function all together.

Being lazy, I ignored this until I needed one of these applications.

At current, I’m attempting to mount the partitions with specified permissions and ownership.

UUID=4F975E2D139676A9 /media/HOME     ntfs    defaults,umask=077,uid=1000,gid=1000 0       1

The important entries are:

• umask=077-This entry sets the permissions on all NTFS-mounted files and folders such that only the owner has access. This satisfies many applications, such as ssh, that require that files are only accessible to their owner.
• uid=1000-This entry sets the ownership on all NTFS-mounted files and folders such that the owner is me. This solution is only useful because I’m the only one using this machine.
• gid=1000-This entry sets the group for all NTFS-mounted files and folders such that the group is my primary group. This solution is only useful because I’m the only one using this machine.

This allows wine to work and gets rid of messages from gdm. I’m not sure what other applications may object to these permissions being set as such on the home directory. If I want to run anything on files in the home directory that relies on a user other than me (such as a web server), it won’t work.

This interesting set of circumstances actually exists, and quite frequently, evidently.

I recently began running Kubuntu on my laptop and started sharing the NTFS partition between Kubuntu and Windows. Unfortunately, after installing VMWare, I found out that VMWare wouldn’t start any of my virtual machines.

I was quite confused. I’d done this so many times on CentOS, I didn’t really expect Kubuntu/Ubuntu, a more user-friendly Linux distribution, to have any issues.

Well, I was wrong. It turns out that adding

mainMem.useNamedFile = "FALSE"

to the given virtual machine’s .vmx file fixes the problem.

This option forces VMWare to actually use the OS’s swap partition rather than making a swap file. Creating the swap file on NTFS seems to be an issue with NTFS on Kubuntu.

Not all systems are synchronized to the time zone that you live in. In fact, if you rent a server like I do, it may be in an entirely different country.

CentOS has several commands concerning timezones. I don’t find some of them to be overly intuitive and figured that I would outline some easy-use recipes here:

1. Finding your system-wide timezone

The best documented way I’ve found to do this is with the following commands:

/usr/sbin/zdump /etc/localtime

You will get output like the following:

/etc/localtime  Sat Jul  5 16:42:53 2008 PDT

Which shows us that the local system-wide time zone for the server itself is “Pacific Daylight Time”.

Because we can set the time zone for our account; using the date command reveals the current time, but the wrong time zone:

# date
Sat Jul  5 19:42:53 EDT 2008

2. Setting your system-wide time zone

Setting the system-wide time zone can be achieved two ways.

One, you can copy the time zone file from /usr/share/zoneinfo to /etc/localtime.

Two, you can run the system-config-date program and choose the timezone tab. This is by far, the easiest way.

3. Setting your account time zone

Setting the account time zone is achieved by setting the TZ variable to the appropriate variable.
In Bourne Shell or BASH:

export TZ="EST5EDT"

In C-Shell or TCSH:

setenv TZ "EST5EDT"

To find out what the variable should be set to, look for the appropriate file name in /usr/share/zoneinfo or you can run the tzselect command.

4. Finding your account time zone

As noted before, running the date command shows the time zone for your environment: