Learn what is tmpfs, what is the use of tmpfs, what is swap, what is the use of swap and differences between tmpfs, and swap.
tmpfs and swap
On the social share of our last post about RAM disk in Linux we got a comment “what is the difference between RAM disk and SWAP?” So I decided to explain it a bit in an article on our blog. In this post I will try to explain how swap and RAM disk i.e. tmpfs/ramfs is different and how they work.
What is tmpfs?
Tmpfs also mounted as shared memory /dev/shm. tmpfs is a portion of a virtual memory mounted as a file system that helps to speed up applications. It normally is used to transfer data between programs. It appears as a file system but it does not use persistent devices such as a hard disk. Instead it uses virtual memory (a portion of a RAM).
That’s why if you create any file in tmpfs it’s not created on your system disks but in your memory. Whenever you un-mount tmpfs, everything within is lost. Its volatile storage. Even if you add an entry of tmpfs to re-mount at boot, it will be mounted blank. Data does not persist over reboots or shutdowns in tmpfs.
What is SWAP?
swap is a portion of your hard disks used to extend RAM. Its roughly extended RAM by use of persistent storage device. swap only comes in action once your RAM (physical memory) is full. The normal thumb rule is the size of the swap should be double of your physical ram size. But these changes depend on the conditions and system you have. Read how to create extra swap here & check swap on the server.
Even if it uses persistent devices, it still is a volatile memory. It does not hold data over reboot or shutdowns. Since it plays the role of RAM, its characteristics are still of ram even if it uses hard disks.
Difference between tmpfs and swap
tmpfs uses memory while as swap uses persistent storage devices.
tmpfs can be viewed as a file system in df output whereas swap doesn’t
swap has general size recommendations, tmpsfs not. tmpfs size varies on system purpose.
tmpfs makes applications fasters on loaded systems. swap helps the system breathe in-memory full situations.
swap full indicates system heavily loaded, degraded performance, and may crash. tmpfs being full not necessarily means heavy load or prone to crash.
tmpfs is enhancement whereas swap is a must-have feature!
Few observations on why HPUX is losing its grip in the market. How virtualization, cloud, costing effects HPUX existence in the market.
hpux loosing market
Really? HPUX really losing market? Kind of yes. But still it is one of the preferred OS for mission-critical environments like the defense industry because of its reliability and security. HPUX, one of the oldest OS (35 years old) released in 1982, no doubt is one of the best server OS we have seen in decades! Being a core HPUX certified engineer obviously I appreciate it moreover IBM’s AIX and Sun Micro system’s Solaris flavor.
But nowadays I am seeing fewer companies, corporate going for HPUX maybe because of its cost involved. Virtualization, cloud computing, and open-source Linux OS flavors made very cost-effective choices for corporate. High Availability offered by virtualization and cloud computing wins over the reliability of Unix when it comes to making cheap billing. HPUX being a hardware-dependent Os i.e. running only on HP hardware (inexpensive) and also involves costs in OS license itself including prime utilities like glance etc., choosing HPUX is a heavy pocket decision. In today’s world of cost-cutting, obviously this is one of the reasons, companies are not opting for HPUX.
The boom in cloud computing also has an effect on traditional in house data centers and hence on proprietary OS like HPUX. Most of the cloud world is powered by Linux which adversely affects HPUX market share rather altogether AIX, Solaris too. People opting for cloud since they offer cheap services with no maintenance. Even I have seen companies are moving away from HPUX and shifting servers on Linux or cloud for billing numbers.
Another major leap back is HP backed off from new releases of HPUX. It’s like they opt to kill their own lion. The latest HPUX roadmap doesn’t even show HPUX v4 and v5 which was existing in old roadmaps. This clearly indicates they have stopped development for new versions and will be continuing current last HPUXv3 as the latest till 2025. This is a major setback since customers lookout for further development plans of service when opting for it.
I wish HP wake up to new versions and we see HPUX back alive and roaring on the IT world! Still it’s not the end of an era! There are many crucial companies and sectors still run by HPUX cause crucial data needs strong shoulders!
These are my observations and perceptions of HPUX’s current stand-in in the market. Let me know if you have any infographics, new data, matrices, graphs about the HPUX journey to date in comments.
Short tutorial explaining what is RAM disk and how to create a RAM disk in Linux. It also includes differences between ramfs and tmpfs.
RAM disk in Linux
Recently one of our readers asked “how to create a RAM disk in Linux?”. So I thought of writing this small tutorial which will help you to understand what is a RAM disk, what is the use of it and how to create a RAM disk in Linux.
What is RAM disk?
Roughly RAM disk can be termed as a portion of your RAM mounted as a directory. It uses tmpfs or ramfs. Refer below table for the difference between ramfs and tmpfs.
ramfs
tmpfs
Old type
New type and replacing ramfs now
Can not be limited in size
Size limit can be defined
Since can not be limited may lead to system crash
Once limit reached, disk full error written. No system crash issue.
Entry is not visible in ‘df’ output. Need to calculate by using ‘Cached’ number in ‘free’ output.
Can be seen in ‘df’ command output
Work mechanism as file system cache
Work mechanism is as partition of physical disk
RAM disk is a very high speed, high performance and almost zero latency area to store application files. Due to its performance-oriented nature, its mostly used for temporary data like caching application files.
How to create RAM disk?
RAM disk can be created in simple two steps. One is to create a directory on which it should be mounted and the second step is to mount it on that directory using specific FS type. Make sure you have enough free RAM on the system so that portion of it can be used in RAM disk. You can check it using free command.
Lets create directory /mnt/ram_disk and mount RAM disk on it.
In above mount command, -t should be followed by tmpfs or ramfs type. For ramfs, size is the starting size of RAM disk since ramfs have limitless size. Size followed by the name of the disk (of your choice ex. new_ram_disk). You can verify if it mounted properly using df command.
You can see newly created tmpfs of 1GB size is mounted on /mnt/ram_disk (highlighted above).
You can add below entry in /etc/fstab as well to persist it over reboots as well. But keep in mind that data within RAM disk flushes for each reboot since its backed memory is volatile.
Learn networking basics in the Linux server. What is the network route, how to check routes, how to define static and default routes in the Linux system?
Network routes in Linux
Network routes on Linux servers are the paths or gateways packets should follow to reach their destinations. Routes can be configured on interface level and system level. The default route is also known as default gateway is the IP where the packet should discover their path if the current network or interface (from which packet originated on the server) does not know the path for destination IP. This article is about discussing network routes in Linux. It will cover how to define the network route for the network interface, where and how to define the default route or default gateway in Linux etc.
How to check current routes on system?
Current routes on system can be viewed using below commands.
# ip route
default via 172.31.0.1 dev eth0
172.31.0.0/20 dev eth0 proto kernel scope link src 172.31.4.137
# routel
target gateway source proto scope dev tbl
default 172.31.0.1 eth0
172.31.0.0/ 20 172.31.4.137 kernel link eth0
127.0.0.0 broadcast 127.0.0.1 kernel link lo local
127.0.0.0/ 8 local 127.0.0.1 kernel host lo local
----- output clipped -----
You can observe 172.31.0.1 is defined as default gateway on above system.
How to configure static route in Linux?
In RHEL: You need to edit below parameters in the config file /etc/sysconfig/network-scripts/route-ethX.
Learn what is hostname, how to set hostname and how to change hostname in Debian and RedHat based Linux systems.
Learn about hostname in Linux
The hostname is a prime identity of Linux servers in the human world! Obviously, the IP address is the main component to identify the system in the environment. In this article, we are going to see anything and everything about the hostname. We will walk through what is the hostname, how to set hostname, how to change hostname etc. Let’s start with the basics of the hostname.
What is hostname
The hostname is the humanly readable identity of the server. Any server is identified by IP address in the network but to identify easily hostname is also given. Normally FQDN (Fully Qualified Domain Name) is expected for the system but even Domain name (the name before the dot) is also fine for systems under private networks. The hostname can be alpha-numeric
Generally hostname standards to the maximum of 255 bytes long. But normally people prefer to keep it 10-12 characters long so that it’s easy to remember. Kernel variables _POSIX_HOST_NAME_MAX or HOST_NAME_MAX defines your current max limit of hostname. You can get their values using getconf a command like below :
# getconf HOST_NAME_MAX
64
How to set hostname in Linux
A quick command in all-new Linux distros is hostnamectl. Use set-hostname switch and your new hostname as an argument.
In above files, you can only view current hostname (being used by the live kernel) under proc file only. Rest all files are used to lookup or set hostname at boot time. So if you change hostname using hostname command then it won’t reflect in rest files. It will only reflect in the proc file.
You can set the hostname of your choice in /etc/hostname or /etc/sysconfig/network and restart network service to notify kernel about it.
How to change hostname in Linux
The current hostname can be checked by typing hostname command without any argument. The hostname can be changed by simply using hostname command followed by the name of your choice.
Cautions : Do not change hostname on live production systems!
On Suse systems: Edit file /etc/HOSTNAME and add hostname in it. There will be no parameter and value format. Only you have to enter hostname like below :
# cat /etc/HOSTNAME
kerneltalks.com
Change hostname permanently in clone, template VM & cloud clones
If you have a system which is prepared using clone, template from VMware or cloud clone deploy then you should do the following :
Edit file /etc/cloud/cloud.cfg and change parameter 'preserve_hostname' to true. You can do it using one-line sed script as below :
root@kerneltalks # sed --in-place 's/preserve_hostname: false/preserve_hostname: true/' /etc/cloud/cloud.cfg
Also, change DHCP related parameter DHCLIENT_SET_HOSTNAME in file /etc/sysconfig/network/dhcp to no. So that hostname wont be changed by DHCP in the next reboot. Again, you can use one line sed to do that as below :
root@kerneltalks # sed --in-place 's/DHCLIENT_SET_HOSTNAME="yes"/DHCLIENT_SET_HOSTNAME="no"/' /etc/sysconfig/network/dhcp
That’s it. These are two extra steps you need to take on cloud or VM servers.
How to configure FQDN in Linux
Another thing around the hostname is to set FQDN for Linux server i.e. Fully Qualified Domain Name. Generally you should be doing in via DNS in your environment but /etc/hosts always get checked first. So its good practise to define FQDN at /etc/hosts file
Use <IP> <FQDN> <Hostname> format to add/edit entry in /etc/hosts and you are good to go. Sample entry below –
An article explaining how to tune kernel parameters in the Linux system using command or using a configuration file.
Tune kernel parameters in Linux
In this article we will be discussing how to set or tune the kernel parameter in any Linux system. There are many ways you can do it like setting them in their configuration files or using a system control command sysctl.
sysctl command is used to configure kernel parameters at runtime. Your current kernel parameters values can be viewed with -a switch.
In the above output you can see parameters on left and their current value on the right. Parameters are sorted with alphabetical order and both columns in output are delimited with = sign so that you can sort this output easily using this delimiter.
There are a few parameters you can even view using the proc file system. You can cat their respective files and view values.
# cat /proc/sys/kernel/shmmni
2048
In above example we can see shmmni value is set to 2048.
How to tune kernel parameter
To change the kernel parameter you can define it under configuration file /etc/sysctl.conf and it will be applied at the next reboot. You need to define parameter=value format in this file (ex. kernel.shmmni=4096).
Each new line represents a new parameter and value pair. Values in this file will be loaded at the next reboot. If you want to load this file immediately then you can can do it by using sysctl -p command. It will load /etc/sysctl.conf file in kernel. You can even define values with -w switch explained below.
To change the kernel parameter using sysctl, you should use a write switch -w along with parameter and value. In the below example we are changing kernel.shmmni value to 2048.
You can observe previously kernel.shmmni value was 4096, using -w we changed it to 2048. This change is immediate and does not need a reboot to comes in effect.
Learn everything about Syslog in Linux. Its configuration file format, how to restart Syslog, rotation, and how to log Syslog entry manually.
Linux Syslog configuration
One of the most important daemons on Unix or Linux based system is syslogd! It logs many crucial system events by default. Logs written by syslogd are commonly referred to as Syslog. Syslogs are first logs when you want to trace issues with your system. They are the lifeline of sysadmins 🙂
In this article, we will see configuration files for syslogd, different configs and how to apply them. Before we begin to go through the below files which we will be using throughout this article frequently.
/etc/syslog.conf : syslogd configuration file
/var/log/messages : Syslog file
There are three projects on Syslog daemon spawned one after another to enhance the previous project’s functionality. They are: syslog (year 1980), syslog-ng (year 1998) and rsyslog (year 2004). So depending on which project’s fruit is running on your server, your daemon name changes. The rest of the configuration remains pretty close similar.
Syslog uses port TCP 514 for communication.
syslogd daemon
This daemon starts with systems and runs in the background all the time, capturing system events and logging them in Syslog. It can be started, stop, restart like other services operations in Linux. You need to check which Syslog version (three projects as stated above) is running (ps -ef |grep syslog) and accordingly, use the daemon name.
# service rsyslog status
rsyslogd (pid 999) is running...
# service rsyslog restart
Shutting down system logger: [ OK ]
Starting system logger: [ OK ]
After making any changes in the configuration file you need to restart syslogd in order to take these new changes in effect.
syslog configuration file
As stated above /etc/syslog.conf is a configuration file where you can define when, where, which event to be logged by Syslog daemon. There name changes as per your Syslog version
Here, on the left side column shows services for which you want logs to be logged along with their priority (succeeded by . after service name) and on the right side are actions normally destinations where logs should be written by the daemon.
All the above priorities are given in ascending level of urgency.
Actions/destination :
Those mostly log files or remote Syslog server to which logs get sent. The remote server can be specified by IP or hostname preceded by @ sign.
Syslog
All logs by syslogd are written its Syslog file /var/log/messages. Typical Syslog file looks like :
May 22 02:00:29 server1 rsyslogd: [origin software="rsyslogd" swVersion="5.8.10" x-pid="999" x-info="http://www.rsyslog.com"] exiting on signal 15.
May 22 02:00:29 server1 kernel: imklog 5.8.10, log source = /proc/kmsg started.
May 22 02:00:29 server1 rsyslogd: [origin software="rsyslogd" swVersion="5.8.10" x-pid="1698" x-info="http://www.rsyslog.com"] start
May 22 02:17:43 server1 dhclient[916]: DHCPREQUEST on eth0 to 172.31.0.1 port 67 (xid=0x445faedb)
Here file can be read in below parts from left to right :
Date
Time
Hostname (This is important to identify which server’s log is this on centralized Syslog server)
The service name for which logs were written by the daemon
Separator colon
Actual message or log
The first 5 fields can be used for sorting, filtering logs in various tools, scripts, etc. Since Syslog logs, all events on system, it’s obvious it grows in size pretty quickly. You can manually rotate Syslog over a specific period or you can even use logrotate utility to do it automatically in the background.
Testing Syslog logging
To test if the daemon is logging messages in Syslog or not, you can use logger command. With this command, you can specify numerous options like a priority, service, etc. But even without any argument, you can supply a string to write in Syslog and it will do the job for you.
# logger Writing KERNELTALKS in syslog using logger. Testing...
# cat /var/log/messages |grep -i kerneltalks
May 22 02:31:05 server1 root: Writing KERNELTALKS in syslog using logger. Testing...
In the above example, you can see all entries after logger command are printed in the Syslog file. Since we used logger command and didn’t specify any service, it logged message with userid root in-service field!
Gallery of wannacry ransomware memes that take over Internet after major cyber attack broke over the weekend.
Wannacry ransomware memes
As everyone is aware the Internet world has been hit by ransomware called WannaCry last weekend. It was a nightmare for Wintel admins as their whole weekend was burned into patching windows servers in their environments. Obviously Linux world was pretty quiet since WannaCry targeted Windows operating systems only.
No wonder, the Internet got filled with lots of memes about this cyber attack. I collected a few of them here:
Read & share !
Note: Image credits are links from where I collected that image. The origin/creator of the image might be different.
If you have any meme about wannacry let us know in comments!
Want to have some some fun in Linux? Read on these articles !
Learn how to configure the local APT repository in Debian based Linux systems. Useful article for package management on Linux.
APT repository configuration
APT is package manager that handles Debian packages (.deb). Linux distributions like Ubuntu, Debian uses APT whereas Red Hat, CentOS uses YUM. The package repository is an index of packages that can be used to search, view, install & update packages for Linux In this article, we will be walking through steps to configure the local APT repository.
APT has two types of repositories: complex and simple. We will see a simple repository configuration in this article. For example, we will be keeping two packages in our repository and configure APT to use it. If you know, you can even download packages in .deb format from existing APT repositories! We are keeping our test packages under /usr/mypackages directory. You can choose your own path.
Rest of the process consist of only 3 steps :
Store packages in the designated directory
Scan that directory to create an index
Add index file path to /etc/apt/sources.list
Step 1 :
Store packages in directory (/usr/mypackagesin our case here). I kept below two packages :
# ll /usr/mypackages
total 156
-rw-r--r-- 1 root root 136892 May 17 10:19 python_2.7.11-1_amd64.deb
-rw-r--r-- 1 root root 11064 May 17 10:20 python-tdb_1.3.8-2_amd64.deb
Step 2:
Scan packages directory with command dpkg-scanpackages. This command takes two arguments: first is a directory to scan and the second is override file. For simple repositories, we don’t need an override file so we can use /dev/null as the second argument.
If you get The program 'dpkg-scanpackages' is currently not installed. error then you need to install package dpkg-dev on your server.
# dpkg-scanpackages . /dev/null
Package: python
Source: python-defaults
Version: 2.7.11-1
Architecture: amd64
Maintainer: Ubuntu Developers <ubuntu-devel-discuss@lists.ubuntu.com>
Installed-Size: 635
Pre-Depends: python-minimal (= 2.7.11-1)
Depends: python2.7 (>= 2.7.11-1~), libpython-stdlib (= 2.7.11-1)
Suggests: python-doc (= 2.7.11-1), python-tk (>= 2.7.11-1~)
Conflicts: python-central (<< 0.5.5)
Breaks: update-manager-core (<< 0.200.5-2)
Replaces: python-dev (<< 2.6.5-2)
Provides: python-ctypes, python-email, python-importlib, python-profiler, python-wsgiref
Filename: ./python_2.7.11-1_amd64.deb
Size: 136892
MD5sum: af686bd03f39be3f3cd865d38b44f5bf
SHA1: eb433da2ec863602e32bbf5569ea4065bbc11e5c
SHA256: 5173de04244553455a287145e84535f377e20f0e28b3cec5a24c109e3fa3f088
Section: python
Priority: standard
Multi-Arch: allowed
Homepage: http://www.python.org/
Description: interactive high-level object-oriented language (default version)
Python, the high-level, interactive object oriented language,
includes an extensive class library with lots of goodies for
network programming, system administration, sounds and graphics.
.
This package is a dependency package, which depends on Debian's default
Python version (currently v2.7).
Original-Maintainer: Matthias Klose <doko@debian.org>
Package: python-tdb
Source: tdb
Version: 1.3.8-2
Architecture: amd64
Maintainer: Ubuntu Developers <ubuntu-devel-discuss@lists.ubuntu.com>
Installed-Size: 50
Depends: libtdb1 (= 1.3.8-2), python (<< 2.8), python (>= 2.7~), python:any (>= 2.7.5-5~), libc6 (>= 2.2.5), libpython2.7 (>= 2.7)
Provides: python2.7-tdb
Filename: ./python-tdb_1.3.8-2_amd64.deb
Size: 11064
MD5sum: 05035155e6baf5700a19fb8308beeca1
SHA1: bd9ec7d2a902e6997651efeaa0842bfb4a782862
SHA256: c53fd7dae63a846cc9583c174e1def248f9def2c4208923704f964068f0a5ea5
Section: python
Priority: optional
Homepage: http://tdb.samba.org/
Description: Python bindings for TDB
This is a simple database API. It is modelled after the structure
of GDBM. TDB features, unlike GDBM, multiple writers support with
appropriate locking and transactions.
.
This package contains the Python bindings.
Original-Maintainer: Debian Samba Maintainers <pkg-samba-maint@lists.alioth.debian.org>
dpkg-scanpackages: warning: Packages in archive but missing from override file:
dpkg-scanpackages: warning: python python-tdb
dpkg-scanpackages: info: Wrote 2 entries to output Packages file.
You can see in above output, dpkg-scanpackages checks all packages list their details on terminal. Since command sends output to stdout we will pipe this output with gunzip to create gz index file.
# dpkg-scanpackages . /dev/null | gzip -9c > Packages.gz
dpkg-scanpackages: warning: Packages in archive but missing from override file:
dpkg-scanpackages: warning: python python-tdb
dpkg-scanpackages: info: Wrote 2 entries to output Packages file.
# ll
-rw-r--r-- 1 root root 1130 May 17 10:27 Packages.gz
Now your index file is ready to be used by APT. You need to let APT know that a new index is created and can be used as a new location to scan packages.
Update the APT configuration file /etc/apt/sources.listwith path of the newly created index file. Add below line :
deb file:/usr/mypackages ./
Thats it! Its done. Run apt update to pickup this new repo.
# apt update
Get:1 file:/usr/mypackages ./ InRelease
Ign:1 file:/usr/mypackages ./ InRelease
Get:2 file:/usr/mypackages ./ Release
Err:2 file:/usr/mypackages ./ Release
File not found - /usr/mypackages/./Release (2: No such file or directory)
Hit:3 http://ap-south-1.ec2.archive.ubuntu.com/ubuntu xenial InRelease
Hit:4 http://ap-south-1.ec2.archive.ubuntu.com/ubuntu xenial-updates InRelease
Hit:5 http://ap-south-1.ec2.archive.ubuntu.com/ubuntu xenial-backports InRelease
Get:6 http://security.ubuntu.com/ubuntu xenial-security InRelease [102 kB]
Get:7 http://security.ubuntu.com/ubuntu xenial-security/main amd64 Packages [265 kB]
Reading package lists... Done
E: The repository 'file:/usr/mypackages ./ Release' does not have a Release file.
N: Updating from such a repository can't be done securely, and is therefore disabled by default.
N: See apt-secure(8) manpage for repository creation and user configuration details.
You can see in the above output there are security warnings since e haven’t added release files in our directory. We had configured only simple repo hence we just stick with .deb packages and rest files not included.
Learn the difference between hard links and soft links. Also discover what are they, how to create them, and how to identify them on the system.
Differences between hard link and soft link
One of the frequently asked Linux or Unix interview questions is what is the difference between hard links and soft links? In this post, we will touch base: what is the hard link & soft link, main differences between hard and soft link, how to create a soft link and hard link, a table showing the difference between a hard and soft link, and how to identify the hard link and soft link.
Without much of distraction, lets get started :
What is hard link?
A hard link is a mirror copy of a file in the Linux or Unix system. Having said that, the original file and link file both have the same inodes. Since both share the same inode, hard links can not cross file system boundaries i.e. you can not create a hard link of a file residing in another mount point. Whenever you delete hard links, the original file and its other hard links still exist since they are all mirror copies. It just reduces the link count! Hard links have actual file content.
What is soft link?
A soft link is just a link to a file in Linux or Unix system. For understanding, you can visualize soft link as a “desktop shortcuts” in windows. Since its a link, its inode is different from the file it’s linking to. Soft links can cross file systems. You can create soft links across file systems. If you delete the original file all linked soft links fail. Since it will point to a non-existent file.
Differences between hard link and soft link :
Hard link
Soft link
Its mirror copy of original file
Its link to original file
Link and original file have same inode
Links has different inode than original file
Can not cross file systems
Can be created across file systems
Show data even if original file deleted
Fails if original file deleted
Has full content of original file
Its just points to source file hence contains no data of source
It can not link directories
It can link to directory
Saves your inodes in kernel since it shared same inode as source
One inode is occupied hence decreasing available inodes
Takes up storage space since its a mirror copy
Takes almost no storage since it contains only path of source
How to create hard link?
To create a hard link you need to use command ln followed by source (original filename) and then link name. In the below example, we are creating two hard links link1 and link2 to file testdata.
# cat testdata
This is test file with test data!
# ln testdata link1
# ln testdata link2
# ls -li
total 12
3921 -rw-r--r--. 3 root root 50 May 16 01:16 link1
3921 -rw-r--r--. 3 root root 50 May 16 01:16 link2
3921 -rw-r--r--. 3 root root 50 May 16 01:16 testdata
You can see above we used ln command to create hard links. Following which we listed their inodes with -i option of command ls. You can see, both links are having the same inode (3921) as the original file (see the first column). Also, the size of hard links is the same as the original file since they contain the same data as the source file.
Now we will delete the original file and see if we can still have data of it from link files.
# rm testdata
rm: remove regular file `testdata'? y
# ls -li
total 8
3921 -rw-r--r--. 2 root root 50 May 16 01:16 link1
3921 -rw-r--r--. 2 root root 50 May 16 01:16 link2
# cat link1
This is test file with test data!
Yes. Data still can be fetched from hard links even after deleting the original file since those are mirror copies!
How to create soft link?
For creating soft links, the same ln command can be used but need to specify -s option (soft link). The rest of the command format remains the same.
# ln -s testdata link1
# ln -s testdata link2
# ls -li
total 4
3921 lrwxrwxrwx. 1 root root 8 May 16 01:26 link1 -> testdata
3925 lrwxrwxrwx. 1 root root 8 May 16 01:26 link2 -> testdata
3923 -rw-r--r--. 1 root root 34 May 16 01:25 testdata
In the above example, after creating soft links if you observe inode numbers of soft links are different from the original file. Also, link size is pretty small since they have only path details of the source, not data. Another observation is soft links shows which file they are pointing to in ls output at last column which was not the case in hard links.
Now, we will delete original file and try to access links.
# cat link2
This is test file with test data!
# rm -f testdata
# cat link2
cat: link2: No such file or directory
You can see in the above output, previously we can use link2 properly. After deleting the original file, links are broken and throwing errors when we try to access them!
How to identify hard link and soft link?
From the above examples, you can figure out soft links are easy to identify. Soft links are marked as link files lxxxxxxxxx in special file bit (first column of ll output). They are even displaying pointers and source file names in the last column ofll output (link1 -> testdata).
Hard links are not that pretty straight forward to identify. You need to use inode option -i inls command and then you need to check for duplicate inodes. This is a manual method. You can even use find command with -same file option. It will then scan inodes and list files with the same inodes (i.e. hard links!)
# find /path -xdev -samefile testdata
The above command will scan /path directory and will list all files having the same inode as testdata file. Which means it will list all hard links to testdata file!
