Continuously monitoring open files in real-time

Recently, I wanted to be able to get a list of files that were being opened by a running process.  Searching all over the web, I found a number of solutions, but they all involved using the lsof command.

The lsof command has many, many options, and it allows you to see which files have been opened by a given process. Coming the other direction, it also allows you to see which process has opened a given file. I can see how it would be extremely useful in many different circumstances.

However, my problem was that my process was opening files and closing them almost immediately. In other words, I had no hope of using the lsof command to view open files, because lsof only shows files that are currently open, and by the time lsof would run, the files were already closed again!

I discovered a different way to continuously monitor, in real-time, all of the files that were being opened by a process, regardless of how quickly they were closed:

strace -tt myprog 2> system_calls.txt
grep 'open(' system_calls.txt > opened_files.txt

strace is a command that logs all of the system calls for myprog.  The -tt option includes a timestamp (with microseconds) at the beginning of each line. Each file is opened with a call to "open(", so grepping for this string should give you a list of all files that were opened.

Installing Ubuntu 10.04 (Lynx) x86_64 Server on Dell XPS 630i

Recently, I needed to install Ubuntu on a Dell XPS 630i.  There was one irritating problem: the installation cd would consistently freeze just after selecting "Install Ubuntu" from the main menu, leaving me with a blinking white cursor and the inner turmoil that can only be experienced while wondering whether your computer is actually doing anything

I've never had these kinds of problems installing Ubuntu before, and I wasn't really sure where to start troubleshooting.  A number of forum websites with postings similar to my own situation recommended changing some of the install parameters, such as noapic, nolapic, noacpi, etc.

None of this worked.

I finally found this post on a Dell community forum, which ingeniously suggested to:

1. Install Ubuntu 8.04 (Heron) x86_64 Server


2. Check for updates in the package manager.  Install all UPDATES (NOT distribution upgrade)


3. Restart


4. In a terminal, sudo update-manager --devel-release


5. Check for updates one more time.  THEN click on the button at the top of the package manager window to install the distribution upgrade to arrive at 10.04 (Lynx).

I never would have thought of that.  I followed the post instructions exactly.  Success!  Everything appears to be working just fine.  Here's to you, jakeman66.

Keeping a remote process running after terminal disconnect

Quoting TheOneKEA at http://www.linuxquestions.org/questions/linux-general-1/keeping-a-process-running-after-disconnect-150235/:

nohup is what you want - it's a wrapper that blocks the SIGHUP signal sent to all applications connected to a terminal when that terminal is closed by the shell.

Just ssh into the box and start the command using this syntax:

[user@remoteboxen user]$ nohup /path/to/command arguments &

The man page explains it better.

Rebuilding the VirtualBox Kernel Modules (Ubuntu 9.04)

Any time there is a kernel update, you would do well to rebuild the VirtualBox kernel module to ensure compatibility with your new kernel version. This can be done by executing the following command from the terminal:

sudo /etc/init.d/vboxdrv setup

Installing Fonts in Linux (Ubuntu 9.04)

First, you can find some good free font downloads at http://www.sostars.com.  I downloaded a stencil font called "Ver Army." I unzipped the file, and found a .ttf font file.

I learned how it install it from this page. Here's a summary:

To install Microsoft Windows fonts: sudo apt-get install ttf-mscorefonts-installer
To install Red Hat Liberation fonts: sudo apt-get install ttf-liberation

To install any other kind of font (including the one I downloaded from sostars.com):

1. mkdir ~/.fonts (make a font directory in your home directory if one doesn't exist already)


2. mv ver-army.ttf ~/.fonts (move your ttf file into the .fonts folder)


3. Restart the computer

GNU sed (Stream EDitor)

sed -r 's/\t+/,/g'



sed	invoke the stream editor
-r	use extended regular expressions (similar to using the -E argument for grep). This gives meaning to the '+' character in my regex.
s	tells sed that we are doing a replacement ("substitution") operation
\t+	find occurrences of one or more tab characters
,	replace it with a comma
g	do this substitution for all occurrences of \t+

So, today I had a problem.  A friend needed me to convert a 10 MB data file from tab-separated format to comma-separated format.

"This should take about 2 seconds."

I wasn't on my trusty little laptop (running Ubuntu 9.04 Jaunty Jackalope since March) and was stuck using a lab computer on campus, which was, of course, running Windows XP with no useful utilities whatsoever.  To try to save some time, I tried to do this conversion right on my friend's computer.  We opened the document in MS Word, and tried to do a Find and Replace for tabs, converting them to commas.

Slow.  Killed the program several minutes into the operation.

Next, over to my trusty laptop.  Loaded up jEdit, a handy programming editor that has done well for me in the past.  Tried to do the find and replace.

Also slow.  Killed this about 10 minutes into the operation.  "It really shouldn't be taking this long."  What went wrong?  JEdit was out of memory.  I found that out from the command-line terminal where I launched jEdit.  Hmmm... Maybe some kind of error box would have been nice so I didn't just sit there for 10 minutes wondering. ;)

No more of this garbage.  We're going to the command line.

Always go to the command line.

I already knew about sed, but my memory was a little rusty on the command-line arguments.  After about 10 minutes, I finally found what I was looking for.

Converted the file in about 2 seconds.

Why is it that something that should take 2 seconds always takes 30 minutes?

Shell script for Google search result parsing

This is the shell script I wrote to help me perform the analysis I did for Quest 5.

1. Perform a site:yoursite.edu search in Google, displaying 100 results per page.
2. Save each page (Google will only give you 10 at most) into a folder named yoursite.edu
3. Download the shell script to the directory that contains the yoursite.edu directory.
4. At the command prompt, type:

./google-results-parse yoursite.edu

5. OR, if you named the yoursite.edu directory something different, run this:

./google-results-parse yoursite.edu savedresultsdirectory

6. It will create a "savedresultsdirectory-parsed" directory, which will contain a "domainlist" file and a "pagelinks" directory. The "domainlist" gives the subdomain breakdown of the search results.  The "pagelinks" folder contains files for each subdomain that include all of the search result URLs for that subdomain.

Download the file here.