Limit the size of .log files & the journal

Introduction

NOTE: This page is still a work in progress.

Log files & the systemd journal do the same thing in different ways. They keep a record of everything that happens on your computer system. This makes it possible to understand what is going right & what is going wrong. As an example, if your system had been infiltrated by an ssh attack, this could be verified in the log/journal. So these log files are good for more than tracking troublesome hardware, or driver problems, badly written network manager code or the plethora of other problems that the complex & dynamic GNU/Linux system has to deal with.

These logs are an absolute blessing, as not all systems have them, & any server administrators who does have them would be very grateful as they can be the bread & butter of what they do.

Generally only server administrators have use for logs that go back any length of time. Few users who run distros on their desktop, Notebook, Netbook... machines, need to keep such huge log files, histories going back for many months or even years, on their system, are a waste of space & also makes viewing your log files more cumbersome.

The first topic on this page will briefly cover the systemd journal

The systemd journal has taken the place of log files though it will happily run in parallel with the standard type log files. These are still created & maintained by default in Arch & Manjaro, which means that even if you delete syslog-ng & all of the /var/log/*log files on reboot you will find the log files have been automatically created again, this will change in the future.

The second topic will cover handling log files

This topic will go into far more depth, it will covering the use of the logrotate command, logrotate.conf, the /etc/cron.daily cron.weekly cron.monthly cron.yearly, some ways to run created scripts, & a mention of the crontab method of running a script also. I'll try to make this section accessible to as many people as possible, which means this will be a long page.

The journal & the logs duplicate the same information

You can read the text of the log files in a text editor, or using the cat, more, less & such commands as you would on any other text file. The journal on the other hand requires the journalctl command to be able to access its contents. The following is a good way to read the journal:

sudo journalctl

How to set a maximum size limit for the journal

You need to use your text editor with root priviliges, starting it with sudo will do the job.

With a simple edit of the /etc/systemd/journald.conf you can set the maximum size limit of the /var/log/journal .

• Uncomment the following line & add the size limit that you want. I set mine to 50MB like this:

SystemMaxUse=50M

After a reboot, you should find that your journal is now somewhat less than 50MB (or whatever limit you set), as it needs to have some working space. Mine went to 38M from 391M.

The Journalctl command - a quick reference [1]

Note: Following are few pointers on things
you can do to make using journalctl quicker, 
easier & more effective, on your system.

You don't have to use sudo with journalctl

Add your user to adm group. This gives your <user> full use of the journalctl command. No more need to use sudo. 

Swap "handy" for your username in the following:

# usermod -a -G adm handy

See the whole line of the journalctl output text

You can pipe the output of journalctl to a file or to a text display tool like "More" or "Less", as follows:

$ journalctl -b -p err|less

Doing so, gives you a means of avoiding the truncation of output which some system displays configurations may experience.

Use a ~/.bashrc alias to make this easy

I use the following ~/.bashrc alias:

alias errors="journalctl -b -p err|less"

On entering errors in the Terminal, all errors or worse since the last boot are sent to (piped) to the Terminal based text display tool called Less which wraps the text output of the journalctl command. Apart from anything else, it makes the errors more useful for anyone reading them in the forum!

Type Q (upper or lower case) to close "Less".

Access to full journal containing info from the system & users:

$ journalctl

Live view, shows the last 10 lines of the journal & all content as it happens:

$ journalctl -f

Basic filtering:

Shows all output to the journal since the last boot:

$ journalctl -b

Shows all output with priority level ERROR & worse, since last boot:

$ journalctl -b -p err

Following is the above command with its output sent to a file called -ERRORS in your /home/<user> directory. Having the - at the beginning of the name should cause the file to be shown at the top of the list of files when viewing the contents of your ~/ (/home/<user>) directory. This command makes it easy to copy the contents of the -ERRORS file, & then paste it to the forum. Doing so allows us to display ALL of the command's output instead of only being able to cut & paste the truncated lines from our terminal:

$ Journalctl -b -p err > -ERRORS

Filtering based on time:

Since yesterday:

$ journalctl --since=yesterday

Give a specific time period:

$ journalctl --since=2012-10-15 --until="2011-10-16 23:59:59"

Pick a specific service & time period:

$ journalctl -u httpd --since=00:00 --until=9:30

Point journalctl at specific devices, services, binaries

Look at a specific device:

$ journalctl /dev/sdc

Check on a binary:

$ journalctl /usr/sbin/vpnc

Check on the interlieved output from two specifics:

$ journalctl /usr/sbin/vpnc /usr/sbin/dhclient

Show all systemd units that have been started in your journal:

$ journalctl -F _SYSTEMD_UNIT

You can then interrogate the journal specifying any of those units.

A summation

The Systemd Journal is capable of advanced functions beyond what has been mentioned here. The above is very good food for thought for people that are wondering if they need to be running syslog-ng or the like that creates most of the /var/log/*log files on their systems.

By experimenting with the above commands one can make an informed decision for themselves, though as mentioned at the beginning of the Journal section, Arch & therefore Manjaro still run both the new systemd journal & the old style log file system in parallel. So if you find the /var/log/*log files to be redundant & you want to be rid of them, various methods would be effective.

Currently I'm running my system with syslog-ng (& its dependency) deleted. I deleted all of the log files from the /var/log directory (leaving any that are in their own sub-directories), except for Xorg.0.log , Xorg.0.old , lastlog , btmp & wtmp (pacman.log turned up when pacman was used). The deleted log files came back on reboot, but having since deleted them again they have stayed deleted for multiple reboots.

Managing /var/log/* files

Introducing Logrotate & friends

What is this Logrotate? [2] logrotate is a powerful tool used to manage the log files created by system processes. It can be instructed to automatically compress, rename in a variety of ways, remove logs, to do all of this & more in a way that maximizes the convenience of logs & conserves your system's resources. An enormous amount of control is available to users including running scripts on your rotated files.

A problem I face in trying to make this article about logrotate as simple as possible is that logrotate can be called in so many ways, & these ways are not mutually exclusive.

For example, logrotate can be called to run on a file, or multiple files in any combination or multiple of hourly, daily, weekly, monthly & yearly, via scripts placed in the /etc/ in the already existing directories hourly daily weekly monthly the yearly directory can be added if required. crontab [3] can be used to run logrotate or scripts as complex as a person needs. logrotate can be combined with other tools in anyway that a user can come up with to process these rotated files at any time & frequency.

The scope of this article

That said, much of the power of logrotate is for the benefit of those administering servers & will not be dealt with in the following. Though what we will deal with can be used on more than just our log files. We can use logrotate to backup any other files that we choose. I will expand on this at a later date.

/etc/logrotate.conf & /etc/logrotate.d

The logrotate.conf configuration file largely dictates logrotate's behaviour, it holds global settings, but most of the work that logrotate does is via script files stored in the /etc/logrotate.d directory, which take precedence over the global settings held in logrotate.conf.

Applications such as Apache, MySQL, Cups & others, put scripts into the /etc/logrotate.d directory to manage their log files.

If you manually run the command sudo logrotate, you will be presented with its usage template. logrotate needs you to specify the path to the script that you want it to use, including the logrotate.conf file which one may think due to its name would be automatically read, it is not.

To run logrotate & the logrotate.conf file you use the following command line:

logrotate /etc/logrotate.conf

Can I store & run my script files elsewhere?

A line exists in logrotate.conf that tells logrotate to run all of the scripts that exist in /etc/logrotate.d

include /etc/logrotate.d

We can use the include command in logrotate.conf to add other directories or use another directory instead of logrotate.d if we have reason to. Be careful what you do as there are files placed into the logrotate.d directory by other programs.

My settings in logrotate.conf don't effect all of the .log files?

Script files that are called via the logrotate.conf file take precedence over the global settings in logrotate.conf . That means that if you call a script from logrotate.conf that is located in the /etc/logrotate.d directory, then that script is more powerful than any of the global setting in logrotate.conf .

I use a script /etc.logrotate.d/rotate.logs that is set to work on all *.log files, & it does. The two that don't get rotated are called faillog & lastlog , apart from not having the .log file extension, these two files are not normal log files, they are accessed via terminal commands of the same name.

Can I store my scripts where I want?

Some applications such as Apache cups, drop scripts into /etc/logrotate.d to aid in their own self maintenance. We can use a location of our choosing for these or other scripts if we want. We just have to call its path in the /etc/logrotate.conf file, the same way, as shown in the following example:

include /home/handy/.config/mylogrotate

Apart from adding our own scripts to /etc/logrotate.d (or any other path that we have chosen to include), we can also add scripts into any of the previously mentioned /etc/ cron.hourly cron.daily cron.weekly cron.monthly folders. OR we can add a script into any of these folders that suit our needs that runs the logrotate /etc/logrotate.conf command which will have the logrotate.conf file, direct logrotate to the default /etc/logrotate.d directory where we have our script(s). OR to another directory where we have our script & have included the path in logrotate.conf . whew!

So you can see there are a variety of ways to call logrotate (let alone use it).

Some uses for Logrotate

For example, script block below does the following, listed line by line:

• /var/log/*.log { specifies the file or the files as this example uses a wild card that says all files ending in .log , the { starts the list of commands that will be used on the file(s) just specified.

• daily Here we are saying cycle these commands daily, we can also say weekly, monthly, yearly (or specify other times with crontab)[4].

• size is where we can place a size limit that will cause a file to be rotated. I placed a 1M one megabyte size limit in the example.

• dateext this puts the date of the rotation on the new copy, so it would use this format: <file.name>.log-20130815

• rotate 7 means keep 7 of our daily (in this script) backups, delete the oldest when it would become the 8th.

• compress is obvious, it uses gzip by default & adds a .gz extension to your file, which will make it look like this: <file.name>.log.1.gz you can choose other compression methods, I'm not going into that here.

• delaycompress tells logrotate to compress the newly rotated file in the next cycle. This has advantages in ease of access & also if the file is still being written to by a process after it has been rotated.

• copytruncate this is a great option, as it copies the contents of the file to a new new file <file.name>.log.1 & then deletes the contents of the original file. You can have no permission problems crop up when you do it this way.

• notifempty do nothing if the file is empty, which makes good logical sense.

• missingok if the file does not exist, give no error.

• } this curly bracket closes the block of commands.

/var/log/*.log {
 daily
 size 1M
 dateext
 rotate 7
 compress
 delaycompress
 copytruncate
 notifempty
 missingok
  }

The above script can be used as is, it does not need to be made executable, it just needs to be put somewhere that logrotate will see (in this example) every day.

We can use the above script block as a template, easily removing parts & modifying its relatively simple settings. It can duplicated in a script with each script block specifying custom settings tailored for individual files.

An Example that you can modify to suit

I'll show how I have my system set, you can use the information already given on this page & other available on the web to fine tune your set up to suit your needs (if you have the need anyway).

Firstly - Be sure this file is here /etc/cron.daily/logrotate

#!/bin/sh

# nicenesses range from -20 (most favorable scheduling) to 19 (least favorable)
NICE=19

# 0 for none, 1 for real time, 2 for best-effort, 3 for idle
IONICE_CLASS=2

# 0-7 (for IONICE_CLASS 1 and 2 only), 0=highest, 7=lowest
IONICE_PRIORITY=7

CMD_LOGROTATE="/usr/bin/logrotate /etc/logrotate.conf"

if [ -x /usr/bin/nice ]; then
  CMD_LOGROTATE="/usr/bin/nice -n ${NICE:-19} ${CMD_LOGROTATE}"
fi

if [ -x /usr/bin/ionice ]; then
  CMD_LOGROTATE="/usr/bin/ionice -c ${IONICE_CLASS:-2} -n ${IONICE_PRIORITY:-7} ${CMD_LOGROTATE}"
fi

${CMD_LOGROTATE}

exit 0

Secondly - Create /etc/logrotate.d/rotate.logs using the following

## rotate all /var/log files with names ending in log
/var/log/*log {
## cycle through these commands once per day
 daily
## keep the results of 7 cycles
 rotate 7
## use gzip to compress each rotated (copied) log file
 compress
## compress the file on the next cycle
 delaycompress
## copy the contents of the log file to a new file <name>.log.1
## & then delete the contents of the original log file
 copytruncate
## do nothing to empty files
 notifempty
## create no errors if a file is missing
 missingok
## after the files have been rotated run the following command
  } 

A Summary of the above example thus far

The First step puts a file into /etc/cron.daily which is an easy way to add the script to a daily cron job. Which means that script will be run everyday.

It basically runs this command:

logrotate /etc/logrotate.conf

As logrotate.conf goes through its list of commands it calls this one:

include /etc/logrotate.d

Which means that any scripts that are inside of /etc/logrotate.d are also run.

This brings us to the second step (above), where we created /etc/logrotate.d/rotate.logs . This script will be run everyday. The comments I added to the rotate.logs file above give a general idea of what it does. You can delete, modify & add to that script, but do it carefully.

The effect of running /etc/logrotate.d/rotate.logs everyday

Is that any file in /var/log that had log at the end of its name will be processed by the commands in the rotate.logs script. This will back up these files to a new file <name>.log.1 & empty the original file to size 0. Any previous copies of with <name>.log.<number> will have their numbers bumped up one, until the day when they would have been given an 8, that is the day that they are deleted.

As well as this rotating (copying) & renaming of files, all files will be compressed in gzip format on the next rotation. Which means that you always have the current file & yesterdays file in /var/log in uncompressed format.

No files that are empty will be processed, & a file being missing will throw no errors.

Support

Following is a link to this page's forum counterpart where you can post any related feedback: [5]