Systemd Continued Commanding Linux Systems In 2015

[tinko]

It is hated because it is hard to avoid. It forces people to change distro. There is a sentiment that it is being pushed on users and systemd affects a lot of packages. Maybe systemd is not bad from a technical standpoint, but it's a huge codebase and that's one reason not to like it.

There are other huge codebases that are hard to avoid. Xorg, Linux Kernel, Glibc, GCC... That is hardly an argument.

[jacob]

Can someone explain to me why systemd is hated? I never had any problems with it... OpenRC on the other hand... (but then again I didn't have that many problems with OpenRC either, not so much that I'd bother switching to systemd)

I don't think it's hated. All the major distros have migrated to it and so an absolute majority of Linux users rely on it as of today. I would contend that most of those users don't care and of those who do, most seem perfectly happy with it. It is hated by a very vocal but ultimately small minority.

[Candy]

One viable reason is journalctl and how it stores data, which causes a lot of corruption in the case of a panicked system.

Let me out myself here.

I have serious problems dealing with journalctl, since its introduction into Fedora. Before that I was easily able to deal with the kernel and system text log files in /var/logs. After the introduction of journalctl I wasn't able to get one clear information out of it. To say the truth. I can't really get along with it. I find the output highly irritating and useless.

You often read about people writing "journalctl -b 1" or something. But honestly I still can't get along with it.

I also ran a light modified Fedora system here (1.5 gb of installation). After a few days I usually end up with an x > 100mb binary blobs in the journal log directory. The first thing I ended up is removing the contents of the journal log directory, because it only wastes spaces on my system. The content is usually useless, unreadable and no value at all for either an administrator nor normal user.

And now I ask into the round here...

... who of you can really get along with jorunalctl ? I bet from 100 readers here that got used to syslogd and could easily get all the value information from the old text files, only 20 remain who can use journalctl. The rest simply defends it, skips it, doesn't even care for it or simply like me, can't get along with it.

After all the time I haven't even figured out how journalctl gives output since the last boot. Whatever I tried entering, the output looked always the same unreadable nonreadable crap. I don't even figure whether I get the output from the beginning of the journal file, or from the last reboot, or from the last whatever....

With old syslogd I was able to point midnightcommander over that file, press F4 (which loads VIM here) and easily scroll around it and read all the information that matters for me.

Just my 0.2€ here.

[spstarr]

Let me out myself here.

I have serious problems dealing with journalctl, since its introduction into Fedora. Before that I was easily able to deal with the kernel and system text log files in /var/logs. After the introduction of journalctl I wasn't able to get one clear information out of it. To say the truth. I can't really get along with it. I find the output highly irritating and useless.

You often read about people writing "journalctl -b 1" or something. But honestly I still can't get along with it.

I also ran a light modified Fedora system here (1.5 gb of installation). After a few days I usually end up with an x > 100mb binary blobs in the journal log directory. The first thing I ended up is removing the contents of the journal log directory, because it only wastes spaces on my system. The content is usually useless, unreadable and no value at all for either an administrator nor normal user.

And now I ask into the round here...

... who of you can really get along with jorunalctl ? I bet from 100 readers here that got used to syslogd and could easily get all the value information from the old text files, only 20 remain who can use journalctl. The rest simply defends it, skips it, doesn't even care for it or simply like me, can't get along with it.

After all the time I haven't even figured out how journalctl gives output since the last boot. Whatever I tried entering, the output looked always the same unreadable nonreadable crap. I don't even figure whether I get the output from the beginning of the journal file, or from the last reboot, or from the last whatever....

With old syslogd I was able to point midnightcommander over that file, press F4 (which loads VIM here) and easily scroll around it and read all the information that matters for me.

Just my 0.2€ here.

If systemd dumped journalctl OR provided away to opt out using it then systemd would be perfect. I don't just mean editing systemd-journald config to stop writing the journal I mean allowing me to decouple systemd from managing any logs.

[Krejzi]

Let me out myself here.

I have serious problems dealing with journalctl, since its introduction into Fedora. Before that I was easily able to deal with the kernel and system text log files in /var/logs. After the introduction of journalctl I wasn't able to get one clear information out of it. To say the truth. I can't really get along with it. I find the output highly irritating and useless.

You often read about people writing "journalctl -b 1" or something. But honestly I still can't get along with it.

I also ran a light modified Fedora system here (1.5 gb of installation). After a few days I usually end up with an x > 100mb binary blobs in the journal log directory. The first thing I ended up is removing the contents of the journal log directory, because it only wastes spaces on my system. The content is usually useless, unreadable and no value at all for either an administrator nor normal user.

And now I ask into the round here...

... who of you can really get along with jorunalctl ? I bet from 100 readers here that got used to syslogd and could easily get all the value information from the old text files, only 20 remain who can use journalctl. The rest simply defends it, skips it, doesn't even care for it or simply like me, can't get along with it.

After all the time I haven't even figured out how journalctl gives output since the last boot. Whatever I tried entering, the output looked always the same unreadable nonreadable crap. I don't even figure whether I get the output from the beginning of the journal file, or from the last reboot, or from the last whatever....

With old syslogd I was able to point midnightcommander over that file, press F4 (which loads VIM here) and easily scroll around it and read all the information that matters for me.

Just my 0.2€ here.

I have to find out about why my app crashed at 10 am yesterday on a very busy system that writes a lot to a syslog. What's the easiest way to do that with text logs besides tons of grepping when I can use journalctl --since ...? People hate change, I know ... I couldn't get used to it either in the beginning.

[Candy]

I have to find out about why my app crashed at 10 am yesterday on a very busy system that writes a lot to a syslog. What's the easiest way to do that with text logs besides tons of grepping when I can use journalctl --since ...? People hate change, I know ... I couldn't get used to it either in the beginning.

Well I might be wrong but iirc, most distros had at least two logs. One that stored the kernel messages and one that stored the system messages (including output from userland programs). Whenever I enter journalctl --since OR journalctl whatever, I'd get tons of messages (with a time log of every couple of seconds) from an app called "audit" which stores some random google-chrome stuff inside there - which again eats up most of the space inside this journaldb (making it bigger and bigger).

Now you may imagine that by this, differencing between some kernel output and system messages is getting harder that way...

And with the text logs, they usually store time & date signatures as well. I usually load the text file into VIM, search for 10 am (and your date for yesterday) and my cursor gets me there. Alternative I use grep.

And!

Why change for the sake of changing things ? Just because it's a modern to change everything and irritate people in their workflow ?

Looked at Fedoras mailinglist archives recently ? Change for the sake of changes. Before you had fine Mailman and threaded archives that you can easily browse. Now you need to have a degree in rocket science, because some python nerd changed it into a mess. Just for the sake of changes. Zero benefits for a normal workflow. Only makes things harder. But that's just my opinion.

[Krejzi]

Well I might be wrong but iirc, most distros had at least two logs. One that stored the kernel messages and one that stored the system messages (including output from userland programs). Whenever I enter journalctl --since OR journalctl whatever, I'd get tons of messages (with a time log of every couple of seconds) from an app called "audit" which stores some random google-chrome stuff inside there - which again eats up most of the space inside this journaldb (making it bigger and bigger).

Now you may imagine that by this, differencing between some kernel output and system messages is getting harder that way...

And with the text logs, they usually store time & date signatures as well. I usually load the text file into VIM, search for 10 am (and your date for yesterday) and my cursor gets me there. Alternative I use grep.

And!

Why change for the sake of changing things ? Just because it's a modern to change everything and irritate people in their workflow ?

Looked at Fedoras mailinglist archives recently ? Change for the sake of changes. Before you had fine Mailman and threaded archives that you can easily browse. Now you need to have a degree in rocket science, because some python nerd changed it into a mess. Just for the sake of changes. Zero benefits for a normal workflow. Only makes things harder. But that's just my opinion.

Well, I could have a text log of several thousands of lines, and given that there's 10 am each day, it could be painful to find ((ugh, vim - no thanks) and here people bash about systemd, yet vim is the most unnecessary complex thing I've ever seen).

As for the audit, that isn't journals problem. It just captures output from everything. Instead of having shitload of files scatered across the filesystem, where some of them get overwritten across boots (.xsession-errors, Xorg.log anyone?), journald stores them all. Not elegant I must admit, but you can filter output by each app, including the kernel-only log. journalctl has a tons of options, which again need to be learned (again, I know that learning to do a thing that you already knew to do in a different way is hard and sometimes irritating). Think of it as grep, awk, vim (yuck) combined in one and triggered with certain options (booo, against unix philosophy (really, who cares? not me)).

[tinko]

Let me out myself here.

I have serious problems dealing with journalctl, since its introduction into Fedora. Before that I was easily able to deal with the kernel and system text log files in /var/logs. After the introduction of journalctl I wasn't able to get one clear information out of it. To say the truth. I can't really get along with it. I find the output highly irritating and useless.

You often read about people writing "journalctl -b 1" or something. But honestly I still can't get along with it.

I also ran a light modified Fedora system here (1.5 gb of installation). After a few days I usually end up with an x > 100mb binary blobs in the journal log directory. The first thing I ended up is removing the contents of the journal log directory, because it only wastes spaces on my system. The content is usually useless, unreadable and no value at all for either an administrator nor normal user.

And now I ask into the round here...

... who of you can really get along with jorunalctl ? I bet from 100 readers here that got used to syslogd and could easily get all the value information from the old text files, only 20 remain who can use journalctl. The rest simply defends it, skips it, doesn't even care for it or simply like me, can't get along with it.

After all the time I haven't even figured out how journalctl gives output since the last boot. Whatever I tried entering, the output looked always the same unreadable nonreadable crap. I don't even figure whether I get the output from the beginning of the journal file, or from the last reboot, or from the last whatever....

With old syslogd I was able to point midnightcommander over that file, press F4 (which loads VIM here) and easily scroll around it and read all the information that matters for me.

Just my 0.2€ here.

First, output from last boot:
journalctl -b -1
(journalctl -b 1 would be the output of the boot numbered "1", so most likely a very old one).

So regarding your question: I used to get along with syslogd (which you can still install by the way) and I can use journalctl much more effectively than syslogd could be used. This is not surprising to me, since journalctl has a ton of useful features. If I want to see what gets logged, no matter for what service, when I try to replicate a failure, I run journalctl -f on some terminal. If I want to know what is the likely cause of the true crash duriing the last boot, I run journalctl -p err -b -1. Mostly I want to know, what's wrong with a certain unit and I run journalctl -u gdm (or some other unit name).

Now I have a question to you: How hard have you tried to figure out how journalctl works? If you just run the command you obviously get all messages in chronological order. The first position that's displayed is probably the early boot stuff of your first boot as logged by the kernel, that might indeed be nonreadable. If you jump to the end you'll just see the usual stuff, that was in your pre-systemd-logfiles... Some wifi-noise, some unimportant failures, new devices appearing and disappearing, logins...

Maybe you should try Poetterings blogpost http://0pointer.de/blog/projects/journalctl.html for a very quick introduction to the possibilities.and then check the (very readable) man page for details and further examples.

EDIT:
just saw your next post.

If you want to get the kernel log just use "journalctl -k".

Also nobody is changing stuff for the sake of changing it. The linked article alone already shows many filtering options that are not available with the traditional text log because all those information is lost when you just log one line of arbitrary text (admittedly with timestamps but other than that this has almost no structure at all). You do realize that the systemd guys are not the ones changing the distros right? They cannot change stuff for the sake of changing it. They develop free software, it gets adopted by distributions. Apparently almost any distro maintainer out there considers those changes useful. Also there are reasons for all of those changes, just listen to any presentation about logging by the main developer and he'll present them to you.

[justmy2cents]

Let me out myself here.

I have serious problems dealing with journalctl, since its introduction into Fedora. Before that I was easily able to deal with the kernel and system text log files in /var/logs. After the introduction of journalctl I wasn't able to get one clear information out of it. To say the truth. I can't really get along with it. I find the output highly irritating and useless.

You often read about people writing "journalctl -b 1" or something. But honestly I still can't get along with it.

I also ran a light modified Fedora system here (1.5 gb of installation). After a few days I usually end up with an x > 100mb binary blobs in the journal log directory. The first thing I ended up is removing the contents of the journal log directory, because it only wastes spaces on my system. The content is usually useless, unreadable and no value at all for either an administrator nor normal user.

And now I ask into the round here...

... who of you can really get along with jorunalctl ? I bet from 100 readers here that got used to syslogd and could easily get all the value information from the old text files, only 20 remain who can use journalctl. The rest simply defends it, skips it, doesn't even care for it or simply like me, can't get along with it.

After all the time I haven't even figured out how journalctl gives output since the last boot. Whatever I tried entering, the output looked always the same unreadable nonreadable crap. I don't even figure whether I get the output from the beginning of the journal file, or from the last reboot, or from the last whatever....

With old syslogd I was able to point midnightcommander over that file, press F4 (which loads VIM here) and easily scroll around it and read all the information that matters for me.

Just my 0.2€ here.

or you could just write

systemctl status servicename --lines=number_of_lines_you_want

and you get output only logs for that service. ever since journald my life is just easier. only cases when i actually resort to journalctl command is when i need specific timeline of some log. journalctl also provides ability to specify additional filterings

i could never go back to syslogd

[kaidenshi]

Can someone explain to me why systemd is hated? I never had any problems with it... OpenRC on the other hand... (but then again I didn't have that many problems with OpenRC either, not so much that I'd bother switching to systemd)

A lot of the dislike I've seen stems from the systemd developers' attitudes that GNU sucks, the Linux kernel sucks, and so they are aggressively pushing their own versions of those things (though I personally doubt the kernel is going anywhere despite the rumors from the anti-systemd crowd). In other words, it's the "not invented here" syndrome on a large scale. And to a certain degree, I actually agree with the protesters: There's really nothing wrong with the current GNU userland, as a whole. There's room for improvement here and there, but overall it's a stable, mature, modular, working system.

I'm also a bit disturbed at how quickly and aggressively systemd has taken over most of the main distros, particularly in Debian's case which by all accounts was a hostile takeover resulting in chaos and strife among the leaders of the OS (I've also heard it was a conspiratorial backroom deal or even a huge financial payoff, though I don't put stock in those rumors myself). Given Debian is upstream of many other popular distros, they all of course had to follow suit, even the behemoth that is Ubuntu (and therefore its own huge collection of derivatives).

Granted, none of this concerns me right now, as I mostly run Slackware, OpenBSD, and lately Alpine Linux. However, I get the feeling there will come a time when we no longer have GNU/Linux, and instead have systemd/Linux, or even just Systemd as a complete OS way, way down the line.