Groesbeek, view of the 'National Liberation Museum 1944-1945' in Groesbeek. © Ton Kersten
Fork me on GitHub
Posts tagged as sysadm

Did you run it through TAttr

2017-08-15 (145) by Ton Kersten, tagged as ansible, sysadm

During my last Ansible training the students needed to create some Ansible templates for them selfs. As I do not want to run a testing template against some, or all, machines under Ansible control I created a small Ansible playbook to test templates.

Read more »

Stupid Fedora

2016-05-26 (144) by Ton Kersten, tagged as sysadm

Yesterday I removed a simple package from my Fedora 23 machine and after that I got the message

error: Failed to initialize NSS library


Searching the interwebs I found out I wasn’t the first, and probably not the last, to run into this problem.

It seems that, one way or another, the DNF package doesn’t know about the dependency it has on SQLite. So, when a package removal requests to remove SQLite, DNF removes it without questions. Ans thus break itself.

But how to fix this? DNF doesn’t work, but RPM doesn’t either, so there is no way to reinstall the SQLite packages.

Tinkering and probing I found this solution:


wget ${url}/sqlite-${ver}.fc23.x86_64.rpm
wget ${url}/sqlite-libs-${ver}.fc23.x86_64.rpm
rpm2cpio sqlite-${ver}.fc23.x86_64.rpm | cpio -idmv
rpm2cpio sqlite-libs-${ver}.fc23.x86_64.rpm | cpio -idmv
cp -Rp usr /
dnf --best --allowerasing install sqlite.x86_64

This downloads the SQLite package and SQLite library packages, extracts them and copies the missing files to their /usr destination. After doing that, DNF and RPM get working again. It could be that I downloaded an older version of the SQLite stuff, so to make sure I have a current version I reinstall SQLite again.

Maybe a good idea to fix that in DNF!

Docker panics

2014-04-14 (140) by Ton Kersten, tagged as sysadm

This morning I was messing around with Docker and I wanted to build me a nice, clean container with Ubuntu in it, to test Ansible thingies. I’ve done that before and everything worked as a charm. Until today.

I have this Dockerfile (I’ve stripped it to the bare bones that still fail):

FROM ubuntu:latest
RUN echo "deb precise main universe" > /etc/apt/sources.list
RUN apt-get -y update
RUN apt-get -y upgrade
RUN apt-get -y install git git-flow
RUN apt-add-repository -y ppa:mozillateam/firefox-next
RUN apt-get install -y firefox

and when I run

docker build .

I end up with a beautiful kernel panic. Whatever I try, panic Nothing in any logfile

I’m running kernel version Linux lynx 3.2.0-60-generic #91-Ubuntu SMP Wed Feb 19 03:54:44 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux which had no problems before.

The Docker version is Docker version 0.10.0, build dc9c28f

Is there anybody out there that knows what’s happening?


Ansible @ Loadays

2014-04-05 (139) by Ton Kersten, tagged as ansible, sysadm

Last Saturday I attended Loadays in Antwerp, Belgium.

After listening to Jan Piet Mens’s talk about Ansible, I was up for it.

At 11:30 sharp, I started my own presentation for an almost packed room. It’s called “Ansible, why and how I use it” and you can find it on SpeackerDeck.

It was a lovely talk, with a very knowledgeable crowd.

Please, have a look at it and if you have any questions, let me know.

Thanks to the crew for organizing such a lovely event, every year.

Photos of the event where taken by Robert Keerse and you can see them at his Google Plus page. Do enjoy!!

For those of you with a strong stomach, the complete presentation is on Youtube as well. Have a look at the Youtube stream

rsync on a not standard port

2013-01-21 (131) by Ton Kersten, tagged as linux, sysadm

Today a colleague asked me to sync some files to a server that is not listening on SSH port 22.

I normally create a configuration entry in my ~/.ssh/config file, like

Host tosync
    Port 1234
    User syncuser

and then command

rsync -va --progress --inplace . tosync:

But this time I didn't want to create the entry in my SSH configuration, because I need this trick in a script. So I started to read the rsync manpage and after some experimenting I found

rsync -va --progress --inplace --rsh='ssh -p1234' .

This syncs the current directory to host on port 1234 as user syncuser.

Resize a partition

2012-10-19 (127) by Ton Kersten, tagged as sysadm

I often have to increase the size of a virtual disk on a virtual machine. But I always seem to forget how to do it. I guess I have done it over a 100 times and I cannot remember exactly how I did it. So this blog entry is to help people on how to do this and as a reminder to myself.

This example is done on a virtual machine with CentOS 6, but it can be done on every Linux. And in the fdisk examples I have left out some of the not to interesting lines.

Oke, here we go:

First, shut down your virtual machine and increase the disk size. Then start your virtual machine and go to the console. Now you have a virtual machine with a new disk size, but the current partition table needs to be adjusted to the new disk size. I know it’s possible with parted, but I always seem to end up on systems where it’s not available. So I just use fdisk.

# fdisk /dev/sda

Now give the p command, which prints the partition table and make a note of the start cylinder of the Linux LVM partition. This is the partition we are going to increase.

Please be very careful This trick only works if the partition you want to resize is at the end of the disk and contains a logical volume type system.

Command (m for help): p

Disk /dev/sda: 17.2 GB, 17179869184 bytes

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *           1          64      512000   83  Linux
/dev/sda2              65        2089    16264192   8e  Linux LVM

Delete the LVM partition (we will recreate this later)

Command (m for help): d
Partition number (1-5): 2

and create a new partition with the original starting point. The starting point should be the same, because all the partitions meta data is at the start of the partition.

Command (m for help): n
Command action
   e   extended
   p   primary partition (1-4)
Partition number (1-4): 2

And fill in the start cylinder of the LVM partition we deleted above

First cylinder (1-25600, default 1): 65
Last cylinder, +cylinders or +size{K,M,G} (65-25600, default 25600):
Using default value 25600

and change the type to Linux LVM

Command (m for help): t
Partition number (1-4): 2
Hex code (type L to list codes): 8e
Changed system type of partition 2 to 8e (Linux LVM)

Command (m for help): p

Disk /dev/sda: 26.8 GB, 26843545600 bytes

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *           2          64      262144   83  Linux
/dev/sda2              65        2089    24567892   8e  Linux LVM

If you agree with the new layout, write it to disk with the w command and quit with q. If it’s not the disk disk with the root volume on it, it could be possible that you can skip the next reboot. Just a partprobe could do the trick.

# reboot

First it’s needed to resize physical volume.

# pvresize /dev/sda2

Make sure you know how much free space you now have

# vgdisplay

Make a note of the “LV Name” of the logical volume you want to resize

# lvdisplay

Resize the logical volume. I use gigabytes as an example here.

# lvresize -L +[Size]GB [LV Name]

Resize the file system on the logical volume.

# resize2fs [LV Name]

git status in the prompt

2012-07-23 (124) by Ton Kersten, tagged as code, git, linux, sysadm

Working with git a lot I decided I needed some git status in my prompt.

I searched the web and some solutions where almost what I wanted and this one by Sebastian Celis came very close.

But it didn't work with my version of zsh, because that didn't seem to understand the =~ operator.

I also think Sebastian makes things over complicated and so I changed some things aroud.

This is what I came up with:

First make sure this code is included in your ~/.zshenv file

    unset __GIT_BRANCH
    unset __GIT_BRANCH_DIRTY

    local st="$(git status 2>/dev/null)"
    if [[ -n "$st" ]]; then
        local -a arr

        if [[ $arr[1] = *Not\ currently\ on\ any\ branch.* ]]

        if [[ $arr[2] = *Your\ branch\ is* ]]
            if [[ $arr[2] = *ahead* ]]
            elif [[ $arr[2] = *diverged* ]]

        if [[ $st = *nothing\ to\ commit* ]]

    if [[ -n "$__GIT_BRANCH" ]]
        local s="("
        case "$__GIT_BRANCH_STATUS"
            ahead)      s+="↑"  ;;
            diverged)   s+="↕"  ;;
            behind)     s+="↓"  ;;
        if [[ "$__GIT_BRANCH_DIRTY" = "1" ]]

        printf " %s%s" "%{${fg[yellow]}%}" $s

and set your prompt to something like this

PS1=$'$C_CYAN%n@%m$(prompt_git_info) $C_WHITE%2~$ $C_OFF'

When I now switch to a directory that is under control of git I get gt status messages in my prompt, like

tonk@mach (master) ~/dir$ git commit -a
[master fca5ac3] Nice, new stuff.
 6 files changed, 88 insertions(+), 12 deletions(-)
tonk@mach (master) ~/.dir$ git status
# On branch master
# Your branch is ahead of 'origin/master' by 1 commit.
nothing to commit (working directory clean)
tonk@mach (master) ~/.dir$

No network on CentOS 6

2012-07-17 (123) by Ton Kersten, tagged as linux, sysadm

When installing a minimal CentOS 6 system, minimal really, really means minimal. After a reboot the network interfaces do not start, so network connectivity is non existing.

Looking into that I noticed that the file /etc/sysconfig/network-scripts/ifcfg-eth0 contained


The lines that mess things up are NM_CONTROLLED=yes meaning the interfaces are managed with NetworkManager, which isn’t actually installed as part of a minimal install. You want a minimal install, you get a minimal install. And ONBOOT=no, meaning "do not start the interface on boot". How stupid is that!

The trick is to run something like system-config-network-tui to set the IP addresses manually, but as you might imagine, that's not installed either.

So you best edit /etc/sysconfig/network-scripts/ifcfg-eth0 by hand and set it to:


The USERCTL=... line is optional: If set to yes it lets non-root users control the interface.

After setting this a service network restart will do the trick.

Finding key codes on Linux

2012-07-04 (121) by Ton Kersten, tagged as code, linux, sysadm

It often happens that I get into a situation where I need to know key codes of pressed keys. On my Mac that's simple. Just use the Key Codes by Many Tricks.

But on Linux I constantly was trying to find out which key produced what.

So I ended up writing a program for that. I started of in the shell, but that ended up being rather tricky and unnecessary complicated. So I redid the whole thing in C.

This is the result

 * Program     : code.c
 * Author      : Ton Kersten

#include <stdio.h>
#include <curses.h>

#define DONE    'q'
#define ESC     0x1b
#define SPC     0x20

char ch;

    printf("Press '%c' to quit!\n\n", DONE);

     * Put the terminal in raw mode, with no echo
    system("stty raw -echo");

     * Print the header
    printf("%4s\t%4s\t%4s\t%4s\r\n", "Char", " Hex", " Oct", " Dec");
    printf("%4s\t%4s\t%4s\t%4s\r\n", "----", "----", "----", "----");

     * Set the initial loop value to something odd
    ch = DONE-1;
    while ( ch != DONE )
    {   ch = getchar();

         * Character read. Display it. Look out for < 0x20
        if ( ch < SPC )
        {   if ( ch == ESC )
            {   /*
                 * Esc. Just say 'Esc'
                        "Esc", ch, ch, ch);
            {   /*
                 * < ' '. Print Control character
                        ch-1+'A', ch, ch, ch);
        {   /*
             * Normal character. Display it normally
                        ch, ch, ch, ch);

     * Put the terminal back to something usefull
    system("stty sane echo");

And this is an example of the output

Press 'q' to quit!

Char     Hex     Oct     Dec
----    ----    ----    ----
Esc     0x1b    0033    0027
O       0x4f    0117    0079
P       0x50    0120    0080
Esc     0x1b    0033    0027
[       0x5b    0133    0091
2       0x32    0062    0050
4       0x34    0064    0052
~       0x7e    0176    0126
q       0x71    0161    0113

Shell tip

2012-07-04 (120) by Ton Kersten, tagged as code, linux, sysadm

During one of my teaching sessions a student asked me if it was possible to find the number of spaces in a variable.

As with all questions in Linux and UNIX the answer is a simple

Of course that's possible. In UNIX and Linux everything is possible.

With some sed or awk this can be done within seconds. But I wanted it done completely within the shell, in this case bash.

This is what I came up with

P="John and Paul and Ringo and George where the Beatles"
R=${P//[! ]/}       # Remove everything that is NOT a space
echo ${#R}          # Show the number of characters (spaces) that are left

And this also works in the Korn shell (ksh) and the Z-shell (zsh).