Take note of the GNU date output that appears in “blue color” text (will be explained later):

GNU Date: Convert date to Unix Epoch time / Julian time

UK, London, BST time zone at this time of writing:

walker@walkernews.net # date -d "2008-08-08 20:08:08" +%s
1218222488
walker@walkernews.net # date -u -d "2008-08-08 20:08:08" +%s
1218226088
walker@walkernews.net # date -d "2008-08-08 20:08:08 UTC" +%s
1218226088

US, Utah, MTD time zone at this time of writing:

walker@walkernews.net # date -d "2008-08-08 20:08:08" +%s
1218247688
walker@walkernews.net # date -u -d "2008-08-08 20:08:08" +%s
1218226088
walker@walkernews.net # date -d "2008-08-08 20:08:08 UTC" +%s
1218226088

GNU Date: Convert Unix Epoch time / Julian time to date

UK, London, BST time zone at this time of writing:

walker@walkernews.net # date -u -d "1970-01-01 1218226088 secs"
Fri Aug  8 20:08:08 UTC 2008
walker@walkernews.net # date -d "UTC 1970-01-01 1218226088 secs"
Fri Aug  8 21:08:08 BST 2008
walker@walkernews.net # date -d "UTC 1970-01-01 1218226088 secs" +"%Y-%m-%d %H:%M:%S %z"
2008-08-08 21:08:08 +0100
walker@walkernews.net # date -d "1970-01-01 1218226088 secs"
Fri Aug  8 20:08:08 BST 2008
walker@walkernews.net # date -d "1970-01-01 1218222488 secs"
Fri Aug  8 19:08:08 BST 2008
walker@walkernews.net # date -u -d "1970-01-01 1218222488 secs"
Fri Aug  8 19:08:08 UTC 2008

US, Utah, with MTD time zone at this time of writing:

walker@walkernews.net # date -u -d "1970-01-01 1218226088 secs"
Fri Aug  8 20:08:08 UTC 2008
walker@walkernews.net # date -d "UTC 1970-01-01 1218226088 secs"
Fri Aug  8 14:08:08 MDT 2008
walker@walkernews.net # date -d "UTC 1970-01-01 1218226088 secs" +"%Y-%m-%d %H:%M:%S %z"
2008-08-08 14:08:08 -0600
walker@walkernews.net # date -d "1970-01-01 1218226088 secs"
Fri Aug  8 21:08:08 MDT 2008
walker@walkernews.net # date -d "1970-01-01 1218247688 secs"
Sat Aug  9 03:08:08 MDT 2008
walker@walkernews.net # date -u -d "1970-01-01 1218247688 secs"
Sat Aug  9 02:08:08 UTC 2008

Have the GNU date to display/output consistent Unix Epoch time at different time zones

So, the question now is about how to have GNU date directly output a portable and consistent Julian-to-date or date-to-Julian conversion, especially when date calculation shell script is running on multiple servers that distributed at different time zones?

As you can see from the GNU date examples with “blue text” of output, the -u can directly produces a portable and consistent result of:

• Unix Epoch time/Julian time of the given date “8 Aug 2008 8:08PM and 8 seconds”
• Date converted from the Julian time of “8 Aug 2008 8:08PM and 8 seconds”

Although specify UTC in the date string given to -d option switch produces the same Unix Epoch time/Julian time, the Julian-to-date conversion results are not consistent.

There is nothing wrong with GNU date. It’s not a GNU date bug that causes it to produce different result at different time zones. Indeed, it’s the behaviour of the GNU date option switches!

For example, when the -u is not used, GNU date produces output that requires additional conversion from the given time zones offset.

So, the output of

date -d "UTC 1970-01-01 1218226088 secs" +"%Y-%m-%d %H:%M:%S %z"

indicates the converted time is displayed as “21:08:08 +0100″, meaning that the “correct” time should be “21:08:08 minus 1 hour”.

In Linux command prompt, type info date and browse to the GNU date examples section, the difference between -u and UTC is mentioned.

You can also access to these examples at the official GNU Coreutils online documentation.

Verdict

Specify the -u option switch when executing GNU date for “date to Unix Epoch time” or “Unix Epoch time to date” conversion, e.g.

date -u -d "2008-12-25 12:25:08" +%s
date -u -d "1970-01-01 1230207908 secs"

These two GNU date commands generate same output in UTC time, regardless of the local time zone where the GNU date is executed.

First of all, GNU date is one of the many Linux programs that grouped in a great package, with a given name GNU Coreutils.

GNU Coreutils is bundled with Red Hat Linux distribution, and is believed being a ubiquitous GNU package among Linux distributions.

For example, in my RHEL4 machine, ls -laSh /bin/date shows it’s only 47KB in size and the rpm -qf /bin/date shows it’s part of the coreutils-5.2.1-31 package.

This GNU date can either set or change the current system date/time or to display date/time format according to user queries. For those serious programmers who use Julian time, you should happy to know that GNU date understands UNIX Epoch time too!

What is that Epoch Time exactly?

The computer times are represented as a count of seconds since an epoch, i.e. a well-defined point of time (hence, it’s called Epoch time).

On GNU and POSIX systems, the epoch time is defined as 1970-01-01 00:00:00 UTC. Thus, it’s also known as UNIX time, POSIX time, or some simply calls it Julian time.

Here are some GNU date arithmetic and time calculation examples, which I’ve referred to the official GNU date documentation (also appears in info date):

walker@walkernews.net # date
Thu Aug 28 17:10:38 BST 2008
walker@walkernews.net # date -d "3 hours ago"
Thu Aug 28 14:10:41 BST 2008
walker@walkernews.net # date -d "+3 hours"
Thu Aug 28 20:12:01 BST 2008
walker@walkernews.net # date -d "last month"
Mon Jul 28 17:13:11 BST 2008
walker@walkernews.net # date -d "next month"
Sun Sep 28 17:14:15 BST 2008
walker@walkernews.net # date -d "+2 months"
Tue Oct 28 17:14:56 GMT 2008
walker@walkernews.net # date -d "2 months ago"
Sat Jun 28 17:16:44 BST 2008
walker@walkernews.net # date -d "yesterday"
Wed Aug 27 17:13:35 BST 2008
walker@walkernews.net # date -d "tomorrow"
Fri Aug 29 17:13:55 BST 2008
walker@walkernews.net # date -d "3 days ago"
Mon Aug 25 17:22:42 BST 2008
walker@walkernews.net # date -d "+3 days"
Sun Aug 31 17:23:08 BST 2008
walker@walkernews.net # date -d "last friday"
Fri Aug 22 00:00:00 BST 2008
walker@walkernews.net # date -d "2008-08-22 +3 days"
Mon Aug 25 00:00:00 BST 2008
walker@walkernews.net # date -d "2008-08-22 -5 days"
Sun Aug 17 00:00:00 BST 2008

How to format the GNU date output format? For example, to get the day, weekday, month, year of 23 days ago can be coded as such:

walker@walkernews.net # date -d "23 days ago"
Tue Aug  5 18:23:24 BST 2008
walker@walkernews.net # date -d "23 days ago" +%e
 5
walker@walkernews.net # date -d "23 days ago" +%d
05
walker@walkernews.net # date -d "23 days ago" +%m
08
walker@walkernews.net # date -d "23 days ago" +%-m
8
walker@walkernews.net # date -d "23 days ago" +%b
Aug
walker@walkernews.net # date -d "23 days ago" +%B
August
walker@walkernews.net # date -d "23 days ago" +"%e-%-m-%Y, %B"
 5-8-2008, August
walker@walkernews.net # date -d "23 days ago" +"%e-%m-%Y, %B"
 5-08-2008, August

You may refer to man date for detail list of GNU date output format modifiers. Also, take note the difference of +%m and +%-m is that the later remove the zero padding.

The zero-padding looks neatly when display in string format. However, zero-padded numeric is not valid in Bash shell script arithmetic interpretation, i.e. 08 - 05 will fail as the valid syntax is 8 - 5.

Unix Epoch time calculation examples with GNU date

The +%s is another GNU date output modifier, i.e. to output number of seconds since Unix Epoch. And, the date/time format for -d option switch is close to ISO 8601 date/time format, i.e. YYYY-MM-DD HH:MM:SS (it’s not a strict ISO 8601 Date/Time format, but close to it).

For example, the UNIX time of 25th of December 2008 at 12:25PM 08 seconds is given as this:

walker@walkernews.net # date -u -d "2008-12-25 12:25:08" +%s
1230207908

So, the Julian time of 2008-08-08 20:08:08 (the date and time of Beijing Olympic Opening Ceremony) is as such:

walker@walkernews.net # date -u -d "2008-08-08 20:08:08" +%s
1218226088

That’s 1218226088 seconds elapsed since Unix Epoch time at 2008-08-08 20:08:08.

To convert UNIX Epoch time or Julian time to human readable date/time format:

walker@walkernews.net # date -u -d "1970-01-01 1218226088 secs"
Fri Aug  8 20:08:08 UTC 2008

But, why you need to do so when there are plenty of email clients appear with intuitive GUI (graphical user interface)?

Well, system administrators or programmers who write Linux shell script to send email with attachments of log file, report, music, picture, etc, will find it easier if working with command-line email client (e.g. sendmail or mail program).

How to send email with attachment in Linux command prompt?

Basically, you need two programs for this topic, i.e. an email client (e.g. mail that bundled with most Linux distributions) and a binary file encoder (e.g. uuencode which is bundled with sharutils package).

For example, to write an email message that’s attached with Logo.JPG:

(echo "This is the content of email and it's attached with a jpeg file called Logo.JPG"; uuencode /tmp/1.JPG Logo.JPG) | mail -s "This is email subject" walker@walkernews.net

The parenthesis groups the both echo and uunencode commands to run as subshell processes, where the output of subshell appears as single input to the mail client.

The uuencode encodes the binary picture file (i.e. /tmp/1.JPG) to ASCII format (i.e. Logo.JPG), which is then sent as email attachment.

If you just want to email attachments without writing the content/message, then the command will be as this:

uuencode /tmp/1.JPG Logo.JPG | mail -s "This is email subject" walker@walkernews.net </dev/null

So, it’s your choice, whether to email attachment with content/message or not. But, those two options are available.

The mkisofs package comprises of few tiny but really useful programs to deal with CD image, i.e. isodebug, isodump, isoinfo, isovfy, mkhybrid, and mkisofs.

Out of all these programs, mkisofs is particularly useful in this post, i.e. its ability to efficiently create CD image of directories or filesystem in Linux.

This is not about creating ISO image from a CD-ROM, which can be easily accomplished with dd command (a simple, legacy command in most Linux/Unix system).

Some of the points to dump a directory content into an ISO image format:

1. You could easily burn the ISO image format of directory content onto a CD-ROM
    
2. It’s much efficient to manage the ISO image of directories that contain few thousands of tiny files, either to transfer it over the network (via scp command, ftp protocol, shfs system, samba file sharing, etc) or copy between local filesystem, as there is less overheat of maintaining transfer status for those few thousands files in the session.

Simple example of using mkisofs

For example, to dump the content of DB2 Workgroup v81 directory into a standard CD ISO-9660 filesystem image with file name called DB2Wkg.iso:

mkisofs -o DB2Wkg81.iso DB2Wkg81

To create CD ISO image of directories that contain long file name or non-8.3 format (particularly if you want to burn the CD image for use in Windows system), use the -J option switch that generates Joliet directory records in addition to regular iso9660 file names. For example, to create CD image of Vista SP1 directory:

mkisofs -o VitaSP1.iso -J VistaSP1

In my test, mkisofs takes less than 10 seconds to dump a directory that contains 4870 files totalled to 328MB.

Well, you don’t have to look for any Linux software if you simply want to mount and read ISO image file!

Linux build in loopback device that allows user to easily mount and access CD image files. No complicated installation or configuration is required to use the loopback device.

How to mount an CD image file (ISO format) in Linux?

1. Login with root user ID
    
2. Create a directory to serve as mount point. You may also skip the directory creation by using default directory that created during Linux installation (e.g. /media/cdrom or /mnt/cdrom in some Red Hat Linux editions). Whichever way, make sure the directory mount point is on a Linux file system that has sufficient free disk space for the size of ISO image file. [ My mistake. The Linux mount point doesn't take up space but the CD ISO image is. For example, if the root filesystem left only 100MB free disk space, the /media/cdrom that resides in root filesystem is able to mount a 650MB CD image that kept in /tmp filesystem. ]
    
3. Suppose you want to mount /ISO/rhel4-cd1.iso image file to /media/cdrom directory mount point, execute this mount command

   mount -o loop -t iso9660 /ISO/rhel4-cd1.iso /media/cdrom

4. Now, you can read / access the CD image file that is mounted to /media/cdrom directory, as if you’re accessing the physical CD-ROM loaded into hardware drive.
    
5. To un-mount the loaded CD image file, simply execute umount /media/cdrom or eject /media/cdrom command will do.

Isn’t it the Linux loopback device rocks? This method allows Linux users to easily accessing CD image files at the speed of hard disk access time (which is so much faster than any optical disk drives in the market, even the current Blue-ray technology can’t compete with a SATA hard disk drive in term of access time)!

My testing environment is a machine running on Red Hat Enterprise Linux 4 Update 4, with the bundled VNC Server (i.e. vnc-server-4.0-8.1). Also assume that a Linux user account called “walker” needs the VNC server to start up automatically when Linux boots up.

1. The Linux user account that needs VNC server to automatically start up after system reboot, must have a VNC password. To create a new (or reset a forgotten) VNC password, just login or su (switch user) with that Linux user account and execute this simple command:

   vncpasswd

   Enter a password when prompted, which is used for VNC authentication.

2. A hidden directory named .vnc is created in the user home directory by the vncpasswd command (if it’s not current exists). Execute ls -la $HOME/.vnc command to check if there is a file called xstartup. If this file is not exists, bring up VNC server with another simple command:

   vncserver :1

   If you get this similar message “A VNC server is already running as :1″, meaning that there is another instance of VNC server running with the same display number. To resolve this, just try to replace the :1 with :2, :3, etc. Alternatively, you may execute this netstat command with root user privilege:

   [root@walkernews ~]# netstat -tulpan | grep vnc
   tcp    0    0 0.0.0.0:5801    0.0.0.0:*   LISTEN   3402/Xvnc
   tcp    0    0 0.0.0.0:5802    0.0.0.0:*   LISTEN   8447/Xvnc
   tcp    0    0 0.0.0.0:5901    0.0.0.0:*   LISTEN   3402/Xvnc
   tcp    0    0 0.0.0.0:5902    0.0.0.0:*   LISTEN   8447/Xvnc
   tcp    0    0 0.0.0.0:6001    0.0.0.0:*   LISTEN   3402/Xvnc
   tcp    0    0 0.0.0.0:6002    0.0.0.0:*   LISTEN   8447/Xvnc

   The netstat output shows that there are two VNC servers running with display number 1 and 2. So, for the 3rd VNC server to start, the command should be vncserver :3.
    

3. Edit $HOME/.vnc/xstartup file with your favourite editor, to un-comment these two lines in order to get the “normal” Linux Desktop view:

   unset SESSION_MANAGER
   exec /etc/X11/xinit/xinitrc

4. Switch user to root account (i.e. su - root), edit /etc/sysconfig/vncservers with your favourite editor, append the display number and Linux user account information to the VNCSERVERS (an array variable). This configuration file defines who can start up VNC server with what display number via the VNCSERVERS array (that’s read by Linux start up scripts /etc/init.d/vncserver). For example,

   VNCSERVERS="1:root 2:tester 3:walker"

   That means there are three Linux user accounts (root, tester, and walker) will start up VNC server with display number 1, 2, and 3 respecitively, as Linux boots up.

   Note: Don’t simply add more than one VNCSERVERS array in /etc/sysconfig/vncserver configuration file. Otherwise, only the last VNCSERVERS array will be used.
    

5. Make sure VNC server (the daemon or server process) is set to auto run upon system boots up to your runlevel. For example,

   [root@walkernews ~]# chkconfig ––list | grep vnc
   vncserver  0:off  1:off  2:off  3:off  4:off  5:on  6:off

   The ––list option of chkconfig shows VNC server is set to auto run in Linux runlevel 5 (the default multi-user runlevel with Linux Desktop console). To configure VNC server to auto run when Linux boots into runlevel 5, use the ––level with on option switch:

   chkconfig --level 5 vncserver on

Ok, that’s all you need. You should have the VNC server automatically running when Red Hat Linux boots up at runlevel 5. Although the guide might looks lengthy to you, but works involve shouldn’t take you more than 3 minutes after you get used with Linux!

SSH server supports two modes of authentication where remote SSH client could be authenticated via legacy system login ID and password or with the public-key cryptography method (that provides non-interactive or password-less SSH login capability)

Assumption made for this example:

• System user walker-a at Linux server WalkerNews-A needs a password-less, non-interactive SSH login to walker-b at Linux server WalkerNews-B.
   
• Both Linux servers are running on RHEL 4 update 5 bundled with openssh-server-3.9p1-8.RHEL4.1 and openssh-clients-3.9p1-8.RHEL4.1

At WalkerNews-A Linux command prompt:

• Execute ssh-keygen -t dsa to create a pair of DSA-based public key and private key for SSH public-key cryptography authentication.

  WalkerNews-A [/home/walker-a]$ ssh-keygen -t dsa
  Generating public/private dsa key pair.
  Enter file in which to save the key (/home/walker-a/.ssh/id_dsa):
  Enter passphrase (empty for no passphrase):
  Enter same passphrase again:
  Your identification has been saved in /home/walker-a/.ssh/id_dsa.
  Your public key has been saved in /home/walker-a/.ssh/id_dsa.pub.

  First, ssh-keygen prompts user to specify a path to save the generated key. You can simply accept the default path by pressing ENTER key.

  When comes to enter passphrase (password to decode the keys), again just press ENTER key to accept the default value, i.e. empty for no passphrase. Additional steps are required for password-less SSH login, if a passphrase is entered to secure the keys (I should cover this in next post for clarity).
   

• Now, copy the generated public key (i.e. /home/walker-a/.ssh/id_dsa.pub) to walker-b home directory at WalkerNews-B.

At WalkerNews-B Linux command prompt:

• Create a .ssh (hidden directory) in walker-b home directory (if it’s not currently exist) and make sure the directory access mode set to 700. E.g.

  mkdir .ssh
  chmod 700 .ssh

• Create a text file called authorized_keys in $HOME/.ssh directory (if the file is not currently exist) and make sure this file access mode is restricted to 600. E.g.

  cd $HOME/.ssh
  touch authorized_keys
  chmod 600 authorized_keys

• Append the walker-a’s public key (id_dsa.pub) to the authorized_keys file. E.g.

  cat $HOME/id_dsa.pub >> $HOME/.ssh/authorized_keys

Now, back to walker-a at WalkerNews-A, each of the subsequent ssh or scp connection initiated here to walker-b at WalkerNews-B will be automatically authenticated via public-key cryptography, i.e. without having to interactively enter password.

For example, login as walker-a at WalkerNews-A and execute these at Linux command prompt:

• ssh walker-b@WalkerNews-B will get ssh automatically login as walker-b at WalkerNews-B, without entering a password. (The insecure legacy protocol to accomplish this task is called rlogin)
   
• scp sourcefile walker-b@WalkerNews-B:temp/new.log will get scp automatically transfer sourcefile to walker-b’s $HOME/temp directory and saved with a new file name as new.log. (The insecure legacy protocol to accomplish this task is called rcp)

If you would like to troubleshoot or understand the SSH public-key cryptography authentication processes, specify the verbose option switch (-v) in ssh command:

ssh -v walker-b@WalkerNews-B