Constantly watch (every 2 seconds) which processes consume the most memory (rss) and sort them by the size. Additionally truncate the commands by the current linewidth of our terminal ($COLUMNS, normally 80). Finally the last command shows only the 30 top most mem users.

Memory can just be used or its is free. Thats seems easy :) But if the systems memory is exhausted memory pages can also be swapped to disk. Multiple processes can share the same memory region. Processes can map video ram, files, libraries etc. into their virtual address space VmSize (see: /proc/pid/maps). Identifying how much memory a process really uses might not be as simple as it sounds.

3 Ways to display various memory stats of the system:

Sum up the VmSize from all processes:

Sum up the VmSize from all firefox processes (pids returned by pidof), and print the result every second

Get very detailed memory usage information from all browser processes (here chromium). This can nicely be combined with the sum command above

See which directories take up the most space. The following command does print the size in megabytes (-m) and makes sure du stays in the same filesystem (-x). At the end we sort everything nummeric (-n) by the size

Use top to display all processes matching the string X

Sometimes applications like Java programs spawn new threads like hell. The following command displays a top 10 of current user processes having the highest thread count. -A displays all processes, -o nlwp,pid,cmd chooses the columns we want ps to show. The rest of the command cuts the output (pr), sort ascending by lwp count, combine unique lines (uniq), and head does display the first 10 topmost entrys.

To see the total number of allowed processes on the system do:

It might be important to stop your syslogger cause he will also write to the disk and this may causes a loop. Use dmesg to see which process writes to the disk.

Boot your system with a live Linux from a stick or a CD. Assuming the source system is on drive sdb and the destination drive is sdc the bootet live system is sda. My approach is to just mount both drives into the corresponding source and destination directories and then copy all relevant directories over. Some empty system directories have to be created. The last step is to install the boot loader grub. To make this succeed we have to change our root directory (chroot) to the new copied destination system. Grub needs access to some system relevant directories created by the kernel so we have to mount them before chrooting.

Use Tar to backup the whole system. p does preserve file attributes, S handles sparse files efficiently, --one-file-system makes sure tar stays in the same local file system ommitting directories like /proc. Also I exclude /home here cause its backupped elsewhere. everything is zipped using gzip.

Local copy:

Over network using ssh:

from current dir to users home dir:

Encrypt/Decrypr File With OpenSSL

I suggest using tmux instead of screen, cause screen is not very intuitive to use.

If copy and paste (left click + drag to select middle click to paste) does not work try holding down the shift key for the entire operation

press ctrl+b [ to switch to cursor-movement-mode, press space or in some cases ctrl+space to enter test-selection-mode, press ctrl+w to copy the text, press ctrl+b ] to paste the text

Screen is a full-screen window manager that multiplexes a physical terminal between several processes (typically interactive shells)

Try this wonderful command if your caps lock key does not do what you want

Useful for passwords, filenames and others

If you dont have lorem ipsum

use date and bc to calculate the number of days between: 15 April 2013 and 30 June 2014. The Answer is 441. Date returns the number of seconds %s between, then divide by seconds per day

This uses figlet to display a nicely formatted date and time string in your console window.

Another funny way to display a colorful box with the current date and time inside

The ogg format is considered better than mp3. But sadly some old or cheap devices cant play ogg files. here is how to convert them to mp3 with a bitrate of 80 kbit/s:

sudo apt-get install ffmpeg

I scaled down the video by 1/2, set the crf (Constant Rate Factor) to 32 and tuned for stillimage (which additionally saves some MB)

To do this for lots of files I would recommend to first fix white space errors in filenames. (some tools still not get it) A white space is a separator using it in filenames is just bad.

Often I have the problem that I have a couple of high res images in a directory and want to resize all of them to a smaller size. This example shows a way to rename all files in the directory to a name with a ascending number and then convert all files to a resolution of 800x600 pixels.

To download a certificate on the command line you can use the openssl client to show a certificate. It will also display more information that we cut away using sed. The certificate will be saved in the /tmp directory. echo -n gives a response to the server, so that the connection is released.

Download all .mp4 files and .sha1 checksums from a website

Delete all .svn directories in all subdirectories under the current directory.

Change the access rights of all makefiles in the current and all subdirectories to read write for current user.

This command replaces "searchtext" with "replacetext" in all *.txt files recursive from current directory.

Search in all files (here all mp4 videos) recursive from current directory if its modification date is 2010. If yes print the filename to stdout.

Search recursive and case insensitive in current directory all .txt files. if a file contains the string SEARCH, print the filename.

File will be echoed if file exists and is a symbolic link. see man bash

Usefull if you want to play a random music file. The command does find all files from current directory, counts them and prints a random filename from the list. (not very efficient but it works)

Grep matches if either one foo or bar is present in the line

Create an ext2 filesystem in ram with a size of 8 megabyte

Speed: ultra fast

several ways to create a 10GB sparse file. Sparse files do not actually take up a real size on the disk. It is possible to cerate a 1TB file even if you only have 1GB free disk space. The filesystem only reserves the space for the file.

Speed: very fast

This might be the fastest solution if cerating a file that should take up real disk space (non sparse) but where the content is not important. The file is allocated but no data (e.g. zeroes) are written to the filesystem.

Speed: fast

Allocate a file filled with zeroes (non sparse file). In this example the file contains real zeroes. Generating zeroes is very fast so the speed is comparable to the max writing speed of the disk.

Speed: (very) very slow

Generating good random numbers is very hard. This is why /dev/random is very very slow. A compromise might be using /dev/urandom which is much faster in trade off predictability of random numbers.

Speed: slow

Speed: moderate

If you need multiple numbered files each in a separate directory you can also generate them in parallel with this command (here we create 20 x 60GB random files!)

finally create md5 sums in parallel (optional):

show actual limits: ulimit -a

To Set maximum processes to 2048 you have to edit /etc/security/limits.conf

A hard limit for nproc maximum number of processes should always be set to prevent users to build a fork bomb. A fork bomb can be something like this:

thanks to Alfe (stackoverflow)

if the script is located in /home/timo/script and is executed from / the current working directory $cwd is /home/timo

Use it this way:

or use a function:

Its now the second time I wonder how to correctly break lines in a bash script. Here is a small real life example how it works:

This comes out:

A and C do not use indendation whereas B and E do. D is an error cause bash will interpret this as two commands. So I prefer B or E.

If you get the following message:

If your SSH-Agent sends muiltiple SSH-Keys to the server but none of them matches, the server does no more accept any key after too many keys have been offered. To access the host and fix the problem, e.g. put your key into the ~/.ssh/authorized_keys you can use the password login

Add the -v parameter, to increase verbosity you can also increase the verbosity even more by using -vv, -vvv is the max verbose level

A nice way to troubleshoot SSH issues is to temporary start an own SSH daemon. Here the problem may be isolated and does not interfere with other SSH daemons. Also see why ssh login can still fail

on the server:

on the client:

To raise the log-level of an existing SSH instance edit the following file

Add the line

Restart SSHD and Watch the logfile for any errors while logging in

If you run sshd in the foreground as root login is possible but if running SSH as Daemon the login fails

Check the SELinux status

Solution is to run restorecon

The following command ssh host produced this output

check if both devices are listed

if /dev/urandom or /dev/random or both are missing create them with

Tested with wine-2.15