Love systemd timers(blog.tjll.net) |
Love systemd timers(blog.tjll.net) |
(Yeah, they're pretty useful, especially after you get an LLM to write all the boilerplate for you. The boilerplate was the main reason I preferred crontab before.)
Oh but it won't appear in the timer-specific logs, I guess...
It is succesfully flying 51 year. And will work next 50 years. Systemd probably will changes syntax in next 2 years. Modern development mindset: if tool is not rewritten last month - it is outdated and we need to reinvent it. Probably using blockchanin and AI.
It's pronounced, "primmer."
Am...am I being punk'd...?
I have knocked together a systemd service or three based on google copypasta. But generally, for cron jobs, why make it complicated? One line in /etc/crontab and done. I generally call an encapsulation script that sets the right environment variables, uses absolute paths, captures stdout/stderr if required and so on. I just want the simplest possible way to launch that script on a schedule.
What are people on non-systemd distros like Guix System, Void, PCLinuxOS, and so on using for this? Is there still something better to use than cron?
Admittedly I never learned cron, I use a lot of `sleep` and `countdown` for relative delays instead. Just earlier today I set up a 12h countdown followed by opening a URL with xdg-open since I expect a release around then and don't want to forget. I also threw in a little notify-send command in case my browser isn't visible, I should see that pop up. Considered using espeak, but don't wanna scare myself and/or ruin my watching experience if I'm watching a video at that time.
Instead of `sleep` you can use `at`. But for scheduling `cron` is still the best.
Package: at
at, batch, atq, atrm - queue, examine, or delete jobs for later execution
We have used and still use cron for decades. It does it's job and does it well.
Ain't that the truth. Literally every crontab I've written for the last 10 years has had this in it:
2>&1 | logger -t cron-WHATEVER
...and that does a pretty good job of capturing anything that the script emits and making it easy to grep for in syslog the following morning.
But I'm still amazed at how many crontabs I run across that don't capture any output at all.
It’s… certainly a product of its time. (I have my system mailer set up to actually send mail to my Gmail account, with authenticated SMTP via API keys, which I did 15 years ago and have no recollection of how I even did it. It still works… somehow. I don’t even use Gmail any more, and I’ll be damned if I have to figure out how to do it with fastmail, and lord knows doing unauthenticated old-school SMTP is just gonna get sent to fastmail’s black hole, so that idea ain’t gonna work either.)
More modern doesn't mean any better or more powerful.
I for one hate this need for going "more modern" without having clear and factual advantages.
That is silly.
One of our customers called in with a production down incident caused by a full disk. We got a copy of the VM and took a look. Investigation revealed that / was full because /var/log was full and that our 'logrotate' timer unit that was scheduled to run once a day had run either exactly never or exactly once... I can't remember which. Further investigation revealed no difference in software load or configuration between this VM and a VM that had a functional logrotate timer unit. Exactly one VM out of hundreds of identical VMs at this site (and many multiples of that at other customer's sites) were affected by this. Advising the customer to clear out /var/log and reboot did not unstick 'logrotate', and none of the diagnostics or fixes we could find anywhere unstuck it. Once "systemd-crond" decided to never schedule this job ever again, it stuck to that decision.
After a lot of searching, we found an open bug report from a year or three prior where someone reported exactly the same symptoms and was scheduling a unit with pretty much the same set of unit configuration flags that we were using. The conversation from the core devs ran through the pattern that one gets used to seeing when one runs into SystemD bugs that are caused by extremely complex unanticipated interactions between parts of the project: "That's not a bug, only an idiot would want that to work.", "Oh, we don't document that that's not supposed to work?", "Wow, okay, yeah, I can see how that maybe should work. That it doesn't sure does seem weird.", "Having said that, I don't know if it's supposed to work, or if it's unsupported. Someone should really either document that or fix it."... and then the behavior is neither fixed nor documented. [1] Absent any actual explanation for the failure, we ended up swizzling the options in our 'logrotate' unit and praying that satisfied whatever gremlin arose from the depths to trouble our customer.
SystemD contains an enormous -and ever-growing- amount of accidental complexity, and has a set of core maintainers who are generally disinterested in either documenting the places where one or more complex systems bind together to cause stop-the-world problems or fixing the systems involved so that they don't bind up. It's a fine project until it's very, very suddenly not, and then you're absolutely SOL. If you're lucky, you can shuffle around what you're doing [2] and hope that avoids the problem. [3]
[0] Some folks use the spelling "SystemD" to mock the project. I use the spelling "SystemD" to distinguish between "the entire systemd project" and systemd(1). I do this because some folks will make a claim like "systemd is very, very small and self-contained. I don't understand why anyone would say otherwise.", but what they are actually saying is that systemd(1) is a fairly small program that doesn't do all that much when run as PID 1. It sucks minor amounts of ass that the project and the program it runs as PID 1 share the same name, but what can you do?
[1] No, I don't have a link to the open bug report. This was more than a year ago, so the bug ID has been long forgotten.
[2] The term of art for this practice is "wave a dead chicken at it".
[3] Plus, like, even disregarding most of the rest of my report... how in the hell do you design a cron that knows a job is scheduled to be run periodically, can tell you how long it has been since it last ran, but never manages to run it? To me, that's unforgivable. It's a "You had one job!"-tier cockup.
Now you get to look around the myriad of places where you can put systemd files, and figure out which ones are base services and which ones are custom, with no general convention to go about it. Nope.
/etc/crontab
/etc/cron.d/*
/etc/cron.hourly/*
/etc/cron.daily/*
/etc/cron.weekly/*
/etc/cron.monthly/*
/var/spool/cron/crontabs/*With —-all
>SystemD contains an enormous -and ever-growing- amount of accidental complexity, and has a set of core maintainers who are generally disinterested in either documenting the places where one or more complex systems bind together to cause stop-the-world problems or fixing the systems involved so that they don't bind up.
excellent comment. thx for the long form. im sure it was fueled by excessive frustration.
imagine my surprise to learn that Systemd was causing my long standing frustration with changing my dns settings. and further surprise to learn that server admins have this same issue and many switch away from using systemd-resolved.
That's introductory course to systemd's shenanigans. People are going to tell you that you're not doing it properly, that there's of course this setting (unless that other setting takes precedence etc.), yada, yada, yada.
If I really have to suffer systemd the first thing I do is manually edit /etc/resolv.conf and then chattr +i /etc/resolv.conf.
And of course remove/purge systemd-resolved.
Not only is it "always the DNS" but then things turn from bad to worse when "it's the DNS, but with systemd".
Removing systemd-resolved is the first step. The second one is moving to an OS or a Linux distro that doesn't have systemd at all.
Did you know that it will convert the answer to a relative query [0] that has generated an NXDOMAIN into a REFUSED? It doesn't do this for fully-qualified queries, and it doesn't do this for relative queries that return something other than NXDOMAIN.
Why do they do this... even if ALL of the resolvers configured in resolved return NXDOMAIN for the query? «Because we believe that the standards say that resolvers can return REFUSED for any reason at all. This is any reason. Now get lost.». Why don't they do this for fully-qualified queries? «chirping cricket noises»
[0] I think this might also be known as a "zero dots" query. Assume that your DHCP-provided search domain is home.arpa. You can do 'ping pc' and every resolver I remember using will convert that hostname into 'pc.home.arpa' and do a lookup with that name, rather than the one you entered.
And you have cron on every system including BSD and not sure but probably also AIX/Sun/etc. It is universal and everyone knows about it. If your server doing something weird your probably will check crontab, not some obscure systemd subsystem almost no one knows about.
Instead using standart solutions (POSIX by the way) systemd again found NIH problem with cron and added one more tool in init combain.
As for the rest of the drivel: so what it was used for a long time. That just means it was pretty good. That doesn't mean it has to stay forever, just that a new contender should do things better.
Systemd timers address real shortcomings of cron.
Your argument boils down to: everyone should be stuck with shortcomings of the early days of computing because you don't like new things.
But you are agree cron is pretty good so it seems analogy with flying brick does not work here.
> Your argument boils down to: everyone should be stuck with shortcomings of the early days of computing because you don't like new things.
I would say: do not fix it if it's not broken. You need a really serious reason to change a good software to one with dubious quality and future. And without portability.
> Ambiguous $PATH settings make cron script execution difficult to predict.
What makes you say that? You can set the PATH right in the crontab. Is that harder to "predict" than it being set in /etc/bashrc, ~/.bashrc, ~/.profile, ~/.bash_profile, /etc/systemd/…, or wherever else?
> You might feel cool knowing the scheduling grammar by heart
I've used Linux since 1994 and I don't know it by heart. But luckily it's pre-printed in the crontab as comments:
# For more information see the manual pages of crontab(5) and cron(8)
#
# m h dom mon dow command
The rest of the complaints, sure. Next time I need a cronjob, I'll try it out.
That is not a fair summarization of their point because that is not the grammar. There's commas, slashes, asterisks, combinations, and then if you want randomization you need to put it in the command itself because cron can't do it. (Some crons can, but it's not a general capability of cron.) Writing a non-trivial cron spec is not easy.
This is generally my only real complaint about systemd. I don't care if it is too monolitic, written in C or whatever, I just want a straightforward syntax for straightforward operations. I'd like it if systemd could recognize if a .target file is a shell script and just do "the right thing". Perhaps it would make sense for a timer file to recognize cron syntax as well. Or at least allow for a kind of extensibility so that I can have it supported.
If systemd had a little more respect for existing conventions, I am pretty sure it wouldn't be so controversial. After all, system administrators like it because they use it all the time, but a regular, full-timer user like me, who only deals with it when something is broken or have to use it as a means-to-an-end to set something up, then all friction is annoying and bad UX. (And no, using Nix is not the solution)
When someone inputs something ridiculous like "5,3/4 4-8,11 1 4,5,6,9-11 */2" you get to enjoy the fun of reverse engineering what they meant (it's never what they actually wrote).
And that's before you get to all the extensions supported in some cron environments (but not all).
I find systemd timers a lot more manageable. Things like having control over whether or not long-running jobs are allowed to overlap and the ability to run tasks between start-finish rather than a fixed time window are major improvements for me. At some point my VPS went down because the backup job ran into some kind of symlink loop and cron just kept spawning more and more backup tasks even though none of them finished.
Having to re-write commands and scripts because CRON had its own special PATH was also a pain point, but the same can be true for some types of systemd timers. But: you can execute those timers manually if you want instead of updating the crontab to trigger in 30 seconds and simply waiting.
Is your example (which I agree, looks cryptic) any less cryptic in systemd?
I asked jippity, and it said this:
[Timer]
OnCalendar=*-04,05,06,09,10,11-01 04..08,11:03/4,05:00
OnCalendar=Sun,Tue,Thu,Sat *-04,05,06,09,10,11-* 04..08,11:03/4,05:00
> Things like having control over whether or not long-running jobs are allowed to overlap
With cron that's just prefixing the command with `flock -n <lock>`, but sure the "pick somewhere to put the lock" is probably better with systemd.
> Having to re-write commands and scripts because CRON had its own special PATH
Why? Wouldn't you just put that in the crontab? I don't even see this as different. It's in the cron config or the systemd timer config.
The other improvements you mentioned seem good.
What's so hard about "At 5 minutes past the hour and every 4 minutes, starting at 3 minutes past the hour, at 04:00 AM through 08:59 AM and 11:00 AM, on day 1 of the month, every 2 days of the week, only in April, May, June, and September through November"?
(I used https://crontab.cronhub.io/ to decode it, to be fair)
# Run if at least a day has passed since the last run
# and it isn't the weekend.
def should_run(finished, timestamp, dow, **_):
return dow not in [0, 6] and timestamp - finished >= one_day
(job
'(next-minute-from
(next-hour (range 0 24 2))
'(15))
"my-program")
That does require you to still know what the default environment is, but it is a mostly completely clean environment, without any influence from any shell.
I'd have to concur that I agree this is an advantage of systemd.
Odd. This script
#!/bin/bash
set > /tmp/set.txt
* * * * * $HOME/bin/testCronScript.sh
BASH=/bin/bash
BASHOPTS=<redacted because long>
BASH_ALIASES=()
BASH_ARGC=()
BASH_ARGV=()
BASH_CMDS=()
BASH_LINENO=([0]="0")
BASH_LOADABLES_PATH=/usr/local/lib64/bash:/usr/lib64/bash
BASH_SOURCE=([0]="/home/user/bin/testCronScript.sh")
BASH_VERSINFO=<redacted bash 5.3.x>
BASH_VERSION=<redacted bash 5.3.x>
DIRSTACK=()
EUID=13370
GROUPS=()
HOME=/home/user
HOSTNAME=hostname
HOSTTYPE=x86_64
IFS=$' \t\n'
LANG=en_US.utf8
LOGNAME=user
MACHTYPE=x86_64-pc-linux-gnu
OPTERR=1
OPTIND=1
OSTYPE=linux-gnu
PATH=/usr/bin:/bin:/usr/sbin:/sbin
PPID=1337
PS4='+ '
PWD=/home/user
SHELL=/bin/sh
SHELLOPTS=braceexpand:hashall:interactive-comments
SHLVL=1
TERM=dumb
UID=13370
USER=user
_=/home/user/bin/testCronScript.sh
* * * * * root /home/user/bin/testCronScript.sh
Maybe your distro's default cron environment was bad, and you never bothered to check and unset the badness? I'd be surprised if they were unable to make the default environment for Timer Units to be bad.
OK but I don't want to hardcode $PATH in the crontab just so I can test the cronjob. Barring the hardcode, $PATH is one thing when cron runs and another when you try out the command yourself. systemctl start foo.service starts the command inside with the same environment as when the timer fires so you know it'll work the same.
On the flip side, your cron job will run at the time you specify in the crontab. Your systemd timer, on the other hand, may fire at the specified time (and most of the time, it will), but it can also suddenly stop firing once it has fired on a February 29th and then never fire again, due to logic bugs in systemd, or it may or may not fire when you "restart" the timer unit, due to logic bugs in systemd (that's when it only has OnCalendar, so yes, definitely a bug).
Why would that be different with systemd timers? If my ~/.bashrc adds /opt/foo/bin, that's also not part of the systemd timer's PATH, right?
But I guess you're saying the ability to trigger the systemd timer off-schedule is the difference? Yeah, it's annoying with cron to have to temporarily set the trigger two minutes into the future. :-P
Not sure adding that feature justifies a complete rewrite, but certainly a nice addition.
> due to logic bugs in systemd
Yeah my main gripe with systemd and other Lennartware is the extremely low implementation quality, not necessarily the ideas. Though the idea of killing tmux/screen on logout is downright criminal. And the fd passing nonsense[1] for system services is clearly just the idea of a child that found a tool and is misusing it.
[1] which is an awesome and underused feature (https://blog.habets.se/2025/10/The-strange-webserver-hot-pot...), but completely misapplied by systemd.
- the default value is missing some values you would expect, like /use/local/bin and /usr/sbin for root.
- on some distributions (for example Arch Linux) the man page doesn't even say what the default path is, or recommend setting it.
- if you need to add something to the path for a single script, you either need to wrap it with a call to env, set it in a wrapper script, or set the path before the entry and reset it afterwards
- you can't use ~ or $HOME in the path, you have to write out the full absolute path. Which is particularly annoying for user crontabs.
Sure, it isn't too hard to work around those, but IMO systemd timers are a better experience, especially since the default uses the same path as all your other services.
What do you mean by "you would expect", that doesn't also apply to systemd timers? /opt/foo/bin is not in the path. Would you expect that?
And if this is an objective problem, can we just change the cron default PATH?
> - on some distributions (for example Arch Linux) the man page doesn't even say what the default path is, or recommend setting it.
Send a PR. This doesn't seem like an inherent problem.
> - if you need to add something to the path for a single script, you either need to wrap it with a call to env, set it in a wrapper script, or set the path before the entry and reset it afterwards
Or on the line, right?
* * * * * FOO=bar $HOME/bin/foo.sh
> - you can't use ~ or $HOME in the path, you have to write out the full absolute path. Which is particularly annoying for user crontabs.
This is incorrect. You can definitely use $HOME in user crontabs.
I'm still not seeing something that warrants a rewrite. (except what you did not mention, which is the ability to run "trigger this now" as a missing feature)
Yes, but people don't. I've had to debug other's crontabs many times over the last umpteen years.
Two environments that set the PATH differently won't have the same value set, either way.
Is this about "yes, but cron doesn't let me trigger through its environment, and systemd timers do"?
another benefit is having logs in one place for the job; cron's "send a mail when there is any amount of output text" is just annoying behaviour, but also only place to get the job output unless you redirect it somewhere. Also starting from timer vs just doing systemctl start job.service is the same so easier to debug
other than that the few improvements in how to specify run time have been pretty useful.
For example, setting timer as "persistent" will mean any run "lost" to machine powered off will just be ran next time after boot, so you can have job on your PC that is just "run backup at 2AM" and if you turn it off before that you get the backup done first thing in the morning
There is also both random, and fixed (depending on machine UUID) random delay so avoiding thundering herd problem with backups is also pretty convenient.
There is even option to wake a device for the job if necessary tho the problem of shutdown is left to the user. And picking whether to start counting to next timer from previous one or from the job's end.
What I would like also is to have job summary page ("hey this job was done X times but failed Y times") but that's probably better left to external tooling
> You can set the PATH right in the crontab. Is that harder to "predict" than it being set in /etc/bashrc, ~/.bashrc, ~/.profile, ~/.bash_profile, /etc/systemd/…, or wherever else?
There is* a common trap as the cron PATH is usually just /usr/bin:/bin so anything in /usr/local/bin, or in /sbin won't be there.
There will always be a default. Is systemd timer's default inherently correct for all users any more than cron's is?
I'm just playing devil's advocate here, but why not just change cron, then? Good for the goose is good for the gander?
man 5 crontab
Same here.
We are now considered old and therefore irrelevant. The new generation uses timers and couldn't care less about cron that has served us just fine for decades.
I use cron and my general attitude towards LP and systemd is very similar to the attitude of LP and systemd to us.
I just don't get it. Like is the core sentiment "How dare they address obvious system shortcomings"? Is it "I learned once and how dare you think I'm capable of learning again"? Is it "I want others to suffer the way I did to learn job scheduling"?
cron did a job, but had shortcomings. Systemd addresses many of those shortcomings. One day something else will come along and address the shortcomings of systemd, and no one will care about systemd nostolgia. This is how technology is supposed to work: making progress and fixing the shortcomings of the past generation. It's not a religion, we don't have to maintain the weird old ways from the 80's, your soul won't be saved by cron or corrupted by systemd.
That's true, but most people don't know the numbered manual sections, so they get the docs for the cron table command not the cron table config file.
No `man man`? ;)
Btw this is my repo for the backup automation: https://github.com/gchamon/borg-automated-backups
I've noticed more and more open source projects recommending timers as a deployment method and I think that's great!
I am perfectly happy with projects recommending timers as long as I can ignore them and use cron.
Which makes it nice to distribute a tool for NixOS so that it can lean into systemd instead of as some bolted-on afterthought.
Makes me wonder what you'd do if you were distributing a lifecycle-heavy tool for Linux users in general since systemd isn't ubiquitous.
I use a systemd timer to run a monthly scrub for my btrfs pool. Kinda cool how you can do increasingly useful things like skip the next scheduled event if the user initiates a scrub, do or don't accumulate tasks if you have a monthly task but the machine was offline for 6 months -- or fold them into a single task, etc.
systemd.services.sync-recyclarr = {
serviceConfig.Type = "oneshot";
path = [ pkgs.podman ];
script = ''
podman exec -it recyclarr recyclarr sync radarr
podman exec -it recyclarr recyclarr sync sonarr
'';
};
systemd.timers.sync-recyclarr = {
timerConfig = {
OnCalendar = "daily";
Persistent = true;
Unit = "sync-recyclarr.service";
};
partOf = [ "sync-recyclarr.service" ];
requires = [ "podman-recyclarr.service" ];
wantedBy = [ "timers.target" ];
};I will admit thought, timers are up there in terms of being the clunkiest systemd unit type to use on a regular basis. I get why they're split up into two files and require different start vs enable syntax's, but man sometimes I just want to create a file that runs a script and be done with it.
Why? It's one of those fuzzy and somewhat hard to explain things. The systemd approach just maps more cleanly to my mental model of "how things should work" I guess. And maybe some of it is that I did indeed experience plenty of " Ambiguous $PATH settings make cron script execution difficult to predict" in the past, although it's not just that.
I won't sit here and claim that systemd timers are necessarily better than cron in any universal / objective sense. But they've won me over, for what it's worth.
And in fact I do have a use-case for needing to run something ~5 minutes after the system boots and then every ~12 hours onward from there. It's great that systemd timers has me covered!
Yet there's always something new to learn and actually consider as another useful tool.
Like imagine trying to explain systemd timers and services and unit files to a beginner.
I think it's... easier? Like "systemd is the place where your system manages all the processes it needs to run. Part of those processes can be run on a schedule, or on a timer, and you define them using this simple text file".
It is also easier to debug as every job gets its own log rather than trying to write to system mailer nobody had set up with the job errors
But one feature of systemd I will absolutely stand by is nspawn. It's just beautiful.
/some/shell -l myjob.sh
. ~/.profile && cd /some/where && ./job >>cron.log 2>&1(Aside: I wrote this article early last month but it caught on only just recently. For better or worse, touching a third rail topic like systemd seems like a sure-fire way to elicit strong and numerous reactions both positive and negative.)
It's a shame docker never supported it. I feel like if they had got on board all those years ago there would be broad support across the software ecosystem for it and we wouldn't need half of these complicated iptables rules and proxies and service mesh. It would be a step towards a capability based system.
I am dealing with mostly non systemd system: BSD, Alpine, termux On BSD anacron works well, but I do not why I am always running into problems with the cronie anacron implementation. And it is very hard to debug.
I would really like a simple modern cron/anacron alternative.
Cronicle looked cool but it is node.js, a bit heavy and being replace now by their new product called xyOps anyway.
For what it's worth there are usually web apps popping up that can decipher goofy cron time/date incantations. [1] This one has a git repo in the top right, not my repo. Maybe clone it just in case their site goes away some day.
I wish documentation for tools would explain their abstractions concepts in terms of its primitives.
Great post, thanks!
But now obviously we were so blind and wrong all this time and the only true solution is of course systemd.
https://www.freedesktop.org/software/systemd/man/latest/syst...
To do this at the user level, you can add something like "@hourly anacron -t /path/to/anacrontab -S /path/to/spooldir" to the user's crontab, though I've never tried this.
Many cron implementations have a similar mechanism.
This isn't the same as with systemd timer because timer lets you specify when you want to run your service exactly and will fallback to running when the system comes online. With @hourly I lose this control and multiple machines could potentially trigger backups at the same time, hogging the physical hard drives and the network.
I find it very easy to reason about: a single process maps to a single recurring task.
It can track last execution into a file, yielding durable schedules when the host is offline.
> runs the service as soons as the system is available.
cron has the @reboot option which I use for a few scripts and works great.
Not an option either, because if I reboot two machines and the backup starts in both of them it'll cripple my NAS
Cronie doesn't have a `@reboot` meta-trigger?
If nothing else, maybe it could be some kid's high school science fair project idea.
It sat unused and powered off for a couple of years after he passed, until I needed a color print.
Didn't do anything but hook it up to power and print. Took about 1/5 of a page until all colors were back in action, after that it printed about 20 pages flawlessly.
And I printed a lot of photos, notes, documents, etc
I have an ink jet printer that I like. I don't print very often (average a couple pages per week) but when I do it's a mix of documents and photos. The ink isn't cheap, but the quality seems good and for the amount I print the expense is minor.
I think it was software and not hardware, but for some reason when I had that printer hooked up to my computer and idle for more than a week, it would simply stop printing. I probably could have dug through logs and figured it out, but I instead set up a cron job to print a test page every Monday and Thursday. The test pages would just have something on the top that said something like LOL PRINTER WORKS.
This wasn't actually as wasteful as it sounds; I was taking a boatload of math courses and needed tons of scratch paper in order to do my problems. Since it was scratch paper and would eventually end up in the trash anyway, I would usually prioritize doing my problems on failed prints and/or test prints, and I would usually exhaust those and then use blank paper afterwards.
The problem I have found is that nixos doesn't seem to pickup and run systemd timers and services placed into the ~/.config/systems/user folder and additionally things like WantedBy=default.target have no effect.
So after I restart all my services manually on reboot I agree, systems timers are cool.
a) It is way nicer and you get decent validation at build time
b) A LLM can port units over if the need arises; it’s a very light abstraction around systemd syntax
c) I personally don’t see how I would ever move to another distro :)
For example, why do unit definitions have to be actual files on disk? Then, all of these files are reloaded when the daemon reloads, not just changed ones. But, why couldn't there be an API letting me add units programmatically? (There kind of is but it's constrained/inflexible)
Or, why can't I declare multiple units in the same file? It's really designed around the filesystem instead of abstracting at a different level, which is a choice I don't think is smart. It's not like it follows the unix philosophy though.
As for the format used for unit definitions, I wish TOML had been around so they could have had something sensible...
that and cron always felt fragile too with a lot of quirks and limitations you had to work around instead of being a robust thing from the start.
First it caused lots of issues. And didn't deliver anything significant
But the biggest issue has always been architectural, the way systemd keeps absorbing existing projects, and functionality. That keep adding to the more than 1 million lines of C monolith, that can burden progress in the futre
But as long people can replace any of systemd tool, for a tool they like better, all good
Personally I am now using desktop/server distros without systemd, and there is nothing that I miss, everything works... cuda/llama.cpp/steam/docker...
And commands always have to google them anyway, or find in history...
I have done scheme all my life, which is why I prefer shepherd. Not only is it in a syntax that i can use elsewhere, I get completion in Emacs.
Systemctl is OK, but I really do not like that respect of the utilities.
But that's because I'm old because obviously systemd-* is the only right way and everyone else who see things differently is a pundit.
[Service]
Type=oneshot
WorkingDirectory={{ home }}/current/
Environment=RAILS_ENV=production
ExecStart=/bin/sh -lc "bin/db-backup --verbose"
Over all I think Systemd get way to much criticism. You don't have to use all the parts, but if you care to go through the documentation you'll find interesting features such as journald log-shipping and systemd-machined which can manage containers and VMs.
It is way simpler and cleaner than Docker/Podman IMHO.
Could have been YAML.
Could have been XML.
It would also make it much simpler to make good GUI editors for the files instead of the Notepad approach most unix config files take.
Could have been XML Property Lists.
ducks
If I never recall hacking in ulimit calls in the top of buggy shell scripts for crappy old services that done respect pam_limits it won’t be soon enough.
Yes, I have too much time sometimes... and I agree, I don't like the syntax.
I am not the biggest fan of systemd, but today I will always reach for a systemd timer over cron simply due to the sheer amount of bad experiences I've had with cron. Hours upon hours wasted trying to troubleshoot crons that weren't working due to some stupid obscure issue, having to use dirty hacks to monitor for success or retry failed jobs.
A few years ago I was trying to run a very simple bash script with cron and the script just died halfway through for no reason. Nothing in logs, worked fine when run directly, but in cron it just stopped halfway through a loop. Never figured out the cause, just gave up and used a timer instead, which worked fine. Never touched cron again after that.
The ease and convenience of monitoring and troubleshooting alone are worth switching over.
Once you learn that env in cron is not same as in your shell and once you learn to redirect output to loggers - it works just fine.
It would be a lie to say that I never debugged cron and sure it's annoying.
> and the script just died halfway through for no reason
Unrelated to cron. Bad script.
Oh nooooo... I've been so wrong. Seen the direction most Linux distro took I decided to move to VMs (VMs which runs systemd-less systems), OCI containers (where by definition PID 1 is not systemd) and now an hypervisor to run systemd Linux VM but I'm now into... An hypervisor that is precisely not Linux (so no systemd at all).
I sinned. So now maybe it's time to buy Microsoft stocks, praise Windows [ini] config files, and venere the Linux PID 1 god with its tentacles meddling with every part of the Linux system.
Or not.
Coz it's looks crazy complicated to set them up.
Come on, dude. That's unnecessarily polemic.
cron et al have served us for decades, yes. But that doesn't mean that cron is the solution that needs to accompany us until the heat death of the universe or year 2038, whatever comes first.
I agree, the systemd folks haven't exactly been the best when it comes to PR or when it comes to being even near feature parity with what they tried to replace. But now, they aren't just at feature parity, they surpassed plain old cron.
Maybe it is time to lay cron to rest, at least slowly.
Yeah I agree.
> systemd folks haven't exactly been the best when it comes to PR
It's deeper than that. Systemd folks are enemies of Linux. First, it's "fuck your opinion, do as we say" attitude which makes me want to throw away everything that comes from that poisonous well. Second, it's the embrace and extinguish strategy employed by the systemd project. And third, systemd author is up to no good: https://news.ycombinator.com/item?id=46784572
Nobody's prevented from using cron instead of systemd timers. The significant differences in typical relatively simple cases are ordering:
cron: M H d m Y DOW
systemd: DOW Y-m-d H:M:S [each part optional, with *, *, and 00:00:00 defaults]
And then, because - is taken, ranges use .. in systemd. Aside from that, it's mostly the same for typical cases of simple periodic timers. Even x/n and x-y/n for steps work similarly. Syntax for complex cases start to diverge, for jitter or special numerically-irregular DOW or DOM or multiple non-periodic times.
In your example, adding more spaces between the date and time parts would make it more visually digestible. There's also the .. range operator which jippity strangely didn't use for the month field even though it did for the hours field.
From what I see not better or worse, it's just different.
CPU speeds have increased & and i/o latency has decreased so much since then that startup times are generally imperceptible, so the pendulum has swung back to favouring socket activation.
The anti-systemd "traditionalists" never seem to acknowledge that history, though!
1) It's supported by cronie. I bet it's supported by many other crons.
2) "Great" news! The software in the Systemd Project only officially runs on Linux, so "it's not portable" is a really bad counterargument when "alternatives to some Systemd Project feature" is the discussion topic.
I maintain one of the competitors listed in that README (compose2nix), so I am a bit biased haha.
For now, I prefer the ability to interop with Compose.
The same problems that could be caused by a polluted environment in cron can be caused in reverse by a polluted environment elsewhere, when you unwittingly copy a command that depends on some environment being set. If you are using systemd as the service manager, this necessarily doesn't happen because it's all units. (Well, you could still copy something from outside of systemd and run into a similar problem, but at least there's essentially only one set of caveats you have to learn for whatever thing you want executed in the background.)
So I guess this isn't so much cron vs systemd timers, but more cron + other init and service supervisors vs systemd init in general.
>
> The same problems that could be caused by a polluted environment in cron can be caused in reverse by a polluted environment elsewhere, when you unwittingly copy a command that depends on some environment being set.
I'm confused about what you need this for? Are you running some utility command that needs the same environment provided by the daemon's service file? If so, any competent init system lets you extend upstream-provided service files. In OpenRC:
# tail -n 1000 /etc/*/test-service
==> /etc/conf.d/test-service <==
extra_commands="${extra_commands} maintenance"
OTHER_THING="overriden other-thing"
maintenance () {
ebegin "doing maintenance. IV='$INIT_VAR' OT='$OTHER_THING'"
set > /tmp/maintenance-set.txt
eend 0
}
==> /etc/init.d/test-service <==
#!/sbin/openrc-run
name="test-service daemon"
command=/usr/bin/socat
command_user=nobody:nobody
command_args="UDP-RECVFROM:6666,fork SYSTEM:'/bin/true'"
supervisor=supervise-daemon
extra_commands="rebuild"
INIT_VAR=${INIT_VAR:-"init var"}
OTHER_THING=${OTHER_THING:-"stock other-thing"}
depend() {
use net
}
start_pre() {
set > /tmp/start-set.txt
}
rebuild () {
ebegin "doing rebuild. IV='$INIT_VAR' OT='$OTHER_THING'"
eend 0
}
# /etc/init.d/test-service start
test-service | * Starting test-service daemon ... [ ok ]
# /etc/init.d/test-service maintenance
test-service | * doing maintenance. IV='init var' OT='overriden other-thing' ... [ ok ]
# /etc/init.d/test-service rebuild
test-service | * doing rebuild. IV='init var' OT='overriden other-thing' ... [ ok ]
# pgrep --list-full socat
133705 /usr/bin/socat UDP-RECVFROM:6666,fork SYSTEM:/bin/true
# diff -u0 /tmp/*-set.txt
--- /tmp/maintenance-set.txt 2026-06-02 23:53:19.703048431 -0700
+++ /tmp/start-set.txt 2026-06-02 23:53:15.265094855 -0700
@@ -9 +9 @@
-BASH_LINENO=([0]="410" [1]="0")
+BASH_LINENO=([0]="409" [1]="0")
@@ -11 +11 @@
-BASH_SOURCE=([0]="/etc/init.d/../conf.d/test-service" [1]="/usr/libexec/rc/sh/openrc-run.sh")
+BASH_SOURCE=([0]="/etc/init.d/test-service" [1]="/usr/libexec/rc/sh/openrc-run.sh")
@@ -30 +30 @@
-FUNCNAME=([0]="maintenance" [1]="main")
+FUNCNAME=([0]="start_pre" [1]="main")
@@ -64 +64 @@
-PPID=133712
+PPID=133702
@@ -69 +69 @@
-RC_CMD=maintenance
+RC_CMD=start
@@ -73 +73 @@
-RC_OPENRC_PID=133710
+RC_OPENRC_PID=133700
@@ -75 +75 @@
-RC_RUNSCRIPT_PID=133711
+RC_RUNSCRIPT_PID=133701
@@ -93 +93 @@
-_='doing maintenance. IV='\''init var'\'' OT='\''overriden other-thing'\'''
+_=']'
But. Another thing that confuses me is why you think that SystemD [0] provides anything special here? If you were to create a service file in most any other service manager and start it with cron, you'd get exactly the same environment sanitization as you get for all other services. Given your testimony, I expect that prior to SystemD, you'd have refused to create service files for things like one-off jobs that weren't system services... so why are you okay with it now that you're using SystemD?
[0] I spell it "SystemD" not to mock it -as I understand some do- but to distinguish The Systemd Project from systemd(1). It sucks minor ass that the two share the same name, but what can you do?
This allows them to work well even if years go by between prints. It's a very thoughtful design element.
(They don't survive sitting for months and months unpowered on a shelf very well, but... you'll have that.)
The only bad thing is that it can get messy if you select the wrong paper type and the ink will not be absorbed by the paper, making the rollers dirty. That’s annoying but fixable, and preventable by not selecting the wrong paper type.
https://www.google.com/search?q=systemd+add+units+programmat...
You can do it either via the CLI (systemd-run), or via the D-Bus API (StartTransientUnit).
SystemD has no dearth of programmable options.
I have tried available options including generators.
https://www.freedesktop.org/software/systemd/man/latest/syst...
E.g. in "Section C" the resulting KeyThree is "value 3▵▵▵▵▵▵▵value 3 continued" where each "▵" symbol is a space.
I think most people would expect a single (or no) space here.
I would guess that most software would strip the comments in SectionC or rearrange the output so that it will result in a diff even when nothing in SectionC changed.
So if you edit the file by hand in the same style as shown in the examples, then most editors would not be able to make a minor edit without making a large diff as many sections would be formatted differently.
XML is that wonderful format that gave us vulnerabilities like death by million laughs, up to a certain moment, you could MitM DTDs, and a whole slew of everything-XML stuff back when XML was like AI is today, none of which I miss today.
Oh, and remember times when programmers would argue whether argument order in XML files should be significant or not?
But XML books with their idealized XML future description did give me the same warm fuzzies as some intricate clockwork mechanism to a Victorian geek.
Personally though I'm not really a fan of TOML when the data is highly structured - prefer YAML even a lot of the time for some visual indication. (If I were writing a parser, I have zero doubt I'd rather parse TOML!)
$ sudo systemctl edit --force --full my-scheduled-work@.timer
$ systemctl edit --user --force --full my-scheduled-work@.timerthere's also `systemctl --all list-timers` to view them.
I mean it should include things that are usually on the path, like /usr/local/bin. And for the root user it should include sbin.
I don't really expect /opt/foo/bin to be there, but if you want to have it available by default everywhere then you can just add it to systemd once, rather than having to remember to add it to crontab as well.
> Or on the line, right?
Oh you're right. I forgot cron evaluates the command with the shell (which has its own set of issues...)
> This is incorrect. You can definitely use $HOME in user crontabs.
You can in the command itself, because that is evaluated by a shell, but not when setting an environment variable. From the Ubuntu crontab(5) manpage:
> The value string is not parsed for environmental substitutions or replacement of variables or tilde(~) expansion
Maybe it depends on the cron implementation though. The cronie documentation doesn't say either way.
So changing the default PATH for cron requires a green field rewrite? You would expect /usr/local/bin. cron doesn't include it. It seems like a disagreement on default settings. I would agree with you about what the default should be, but I don't understand how this has anything to do with cron either as code or architecture.
This seems a bit like /sbin being missing in the default PATH for the root user's regular shell being a reason to rewrite bash to systemd shell.
I never claimed that.
It's just one of many reasons why I think systemd timers are easier to use.
Compare it to the alternative of using plain systemd (including command(s) required to enable units).
Also, consider what build-time validation you get prior to starting the unit/timer. Hint: zero.
I do appreciate build-time checking but I think this can be solved at systemd side as a separate tool just as effectively.
But my question was around readability: were you able to understand what the snippet I shared is doing?
Re: build-time checks - but systemd hasn’t done it, and I am also unsure where exactly this verification would even take place given systemd’s configuration model. Unless you’re talking about some kind of language server or IDE integration.
> Unrelated to cron. Bad script
Again, worked fine when run manually, worked fine in a systemd timer. Pretty sure I still have it running today and it continues to work fine without ever failing.
We are all guilty of making bad scripts, bash is a disgusting degenerate language (and I love it). The way we learn to write good scripts is by writing bad scripts in enough amounts to get bitten by all the warts.
One thing I really love about cron, is that if you set up mail on the server (which: you should btw), then cron actually sends emails if it sees anything in stdout and stderr.
I am a dyed in the wool systemd non-believer, but I really do like the timers.
>Not sure adding that feature justifies a complete rewrite, but certainly a nice addition.
If there is a feature that justifies using a completely different tool it's obviously this one.
You should not see a missing killer feature and go directly to the conclusion of rewriting.
Sure, greenfield is fun. So there's that.
Because you can test if systemctl start foo.service works and if it does, you know it'll work when foo.timer triggers it.
> my main gripe with systemd and other Lennartware is the extremely low implementation quality
It doesn't stack up that poorly against other projects but it's unacceptably low for a core system component.
It's a better workflow than the one I mentioned for cron, yes, but it's nothing inherent. I would not call the difference one of being "predictable", though.
The article and thread is full of complaints about cron that are either just missing features that could be added (like this one, trigger on demand), exactly the same ("the time spec is too complex"), or just plain "no you can already do that with cron".
It wasn't until this comment (https://news.ycombinator.com/item?id=48385824) that I saw viable reasons why one would need to start from scratch.
Except, of course, that "green field is more fun than improving things".
> I don't want to hardcode $PATH in the crontab just so I can test the cronjob
How is this different? Any way you slice it, you have to configure it somewhere if you're not lucky enough for the default to be acceptable.
> [Lennartware] doesn't stack up that poorly against other projects but it's unacceptably low for a core system component.
Well put.
It's a workaround for a design flaw. If you forget it you'll notice when it fails in production. There should rarely be a need to change the default path (just specify the full path if necessary), it should just be the same when you test and when it runs by itself. Trigger on demand would of course solve that.
> exactly the same ("the time spec is too complex")
Of course, the timespec complaint is bullshit. Both have manpages that will tell you immediately if you don't have it memorized.
On the other hand, it is a consistently heard argument when debating why the year of the Linux adoption on desktop hasn't happened yet is that there are too many standards, too many cooks. And I kinda agree with that, packaging software for multiple distributions is a hot mess, especially if you're shipping daemons. systemd is at least one worry less, it's a stable API that can be used for all purposes.
> And third, systemd author is up to no good: https://news.ycombinator.com/item?id=46784572
Again, this isn't as black-and-white as it seems on a cursory review.
From the perspective of someone who hates anti-rooting measures on phones, anything moving into the direction of trusted computing is bad.
But from the perspective of someone, say, in Russia, Iran, the US or Germany who might be a journalist or political activist? Suddenly, a way for the OS to attest if the hardware hasn't been manipulated in an evil maid scenario and that any successful attempt of exploiting an OS vulnerability at runtime can be reverted by a simple reboot becomes extremely vulnerable.
My personal opinion, we need strong and good attestation capabilities for a multitude of use cases. But we also need good laws that protect user freedoms, similar to "right to repair" laws we need "right to root" laws that ban applications from requiring unrooted phones.
However absolutely fuck all the metastasis surrounding systemd and the shit inside, timers included.
> we need strong and good
Yeah this is true.
Problem is, this technology will absolutely be used for limiting your access to the internet from a distribution that was modified. Thus leading to locked down personal computers.
Meh. systemctl list-timers is orders of magnitude better than trying to wade through /etc/crontab, /etc/cron.{hourly,daily,monthly,weekly,yearly} and whatever else non-Debian distributions come up with, and cron's arcane syntax is a mess on its own on top of that. Especially if you try to run a cronjob as not-root.
> Problem is, this technology will absolutely be used for limiting your access to the internet from a distribution that was modified. Thus leading to locked down personal computers.
That's a hypothetical threat IMHO. DRM and Cloudflare Turnstile [1] are actual threats that are far bigger in practical usage.
That isn't something I'd want to happen, it sounds like it creates a potential queue of scripts that will flood the system on start, if it works the way you described.
I prefer the deterministic behavior of cron, the script will run when it is specified to run, as you said earlier, as long as the system is running; and as I stated in a separate comment, it will run @reboot if I need it to run then.
> With @hourly I lose this control and multiple machines could potentially trigger backups at the same time
Then don't use @hourly, use staggered times, it's very easy.
There are two options to fix it;
Disable persist so no catching up on missing scripts. Set OnBoot=5m so it gets ran 5 minutes after boot, so your script (say backup) is ran on boot first, then every time on schedule
Enable persist but just add sleep in ExecStartPre - very "cron" way but there is just no in-systemd option to enable "catch up" script to be delayed
Sadly no option to "run catch-up timers with delay" at least yet
> Then don't use @hourly, use staggered times, it's very easy.
Not in cron. In systemd it's just RandomizedOffsetSec=30m and it is "stable" - same host with same job will always have same delay so on multiple hosts it is spread nicely. There is also non-stable version
This isn't what happens. If you leave it offline for days it'll only trigger the service only a single time.
100 jobs all running at different times throughout the week is a very different load than them all falling back and running at the same time on system boot.
https://specifications.freedesktop.org/desktop-entry/latest/...
[Service]
Type = oneshot
WorkingDirectory = %h/current/
Environment = RAILS_ENV=production
ExecStart = /bin/sh -lc "bin/db-backup --verbose"I find myself doing this sort of thing all the time..
Environment = MULTIPLE=environment VARIABLES="in single line"That's what I'm saying though; how is it a design flaw? It's arguably a flaw in the default configuration. That's certainly fixable without changing the design at all.
> Trigger on demand would of course solve that.
Is it a design flaw that you can't trigger the job as a test? No, that's just a missing feature.
> cron did a job, but had shortcomings. Systemd addresses many of those shortcomings
Right, but what are those "many" shortcomings? The article lists four, and fully half of them seem to be nonsense, per my comment. (time spec syntax appears to be equally complex in systemd timers, and I have no idea what they mean about PATH, as it seems equal too)
The remaining two are fairly good points, kind of. Sending mail is a black hole until you look there, sure. Believing that emails get meaningfully delivered on a non-email server is very anachronistic. But isn't logging a black hole until you look there too?
So from the article that leaves "Execution history is difficult to follow and interrogate", which I super agree with. I would argue it's not inherent to cron, and one could have written a small tool that allows following and interrogating.
And… surely that goes for logging too? cron does log to syslog, right?
Maybe there's some integration with other stuff that is better, that I'm not aware of?
I'm not disputing that it's better. I'm sure it is. But where's that list? If it's literally only the `list-timers` command, then that's very underwhelming. Is it the randomness to scheduling? I've never needed it (I just spread them out over an hour by fixed start time), but sure that would have value in other cases where coordination is not possible. You could add it to cron, but not in a nice way.
To me it seems like engineers do what engineers like to do: enjoy greenfield implementations. It's open source. Nobody's going to ding your quarterly performance evaluation for going off on a amusement coding session without a requirements doc. I know that I have written a lot of tooling for amusement and to work the way I want it to work. I certainly understand the engineering mind that would rather write something from scratch than understand the previous system.
I don't mind learning new things, but this article seems to fawn over stuff you can do in systemd timers, where… yeah that was always an option (in cron). I don't have faith that the article writer actually knew cron before they trash it in favor of something else.
On a tangent, I do agree with egorfine that Lennartware inherently has a disdain for users and their workloads (e.g. kills user processes inside a screen/tmux arbitrarily), and that audio on linux was set back 5-10 years just from the mere disastrously bad quality of the pulseaudio implementation. It makes sense that he works for Microsoft.
* coalescing jobs with control over the granularity of it. That means you can say "i want this job run on at 14:30:02 exactly" and I want these jobs run at 19:21 or so, 19:22 or so: and 19:23 and set your resoultion to 10m and they'll all run at once. Great on laptops and other scenarios where you want to reduce power draw.
* System wakeup - you can wake a system from sleep various sleep modes (details depend on hw support) and run a job.
* cohesion with the rest of the system. this is a big deal when you stop playing with just your desktop and pet server and have to deal with 10^4 or more servers. having to deal with the wierd quirks of cron vs inittab vs whatever is frustrating and when there are many people working on it, someone is always going to do something quirky and fragile. Yes you have to know the systemd things, but that's it - a timer starts a unit, any unit, without all the bullshit (e.g, oh im stating with cron, these magic invocations are neeeded, oh im starting it with runit and these different invocations are needed, etc)
I literally never experienced any of the problems people complain about for pulseaudio - at the time it was released it was the smoothest audio experience i ever had on linux. I think some people just want to look cool and complain about the new thing.... But also I read manuals and think for 3 or 4 seconds before doing things, so maybe it has something to do with that.
As for pulse audio: no, it has absolutely nothing to do with that. That's just being condescending and smug. I've never had a lithium battery fire, so that must mean manufacturing defects never cause them?
I'm all for it.
However the trust in systemd brand and its leader is deeply negative, so anything that comes from systemd folks is an immediate no.
I don't live in terror of new things though, so I don't really understand the propaganda.
This is a take that is so drastically different to what I (and many other people) have experienced that it now makes sense that systemd is to your liking.
(Use EnvironmentFile= instead)
In cron, you basically have to either use your configuration management to generate those times, or have a random delay script running before the command
In systemd timers, it's just
OnCalendar=0/6:00:00
RandomizedOffsetSec=60m
If you add
Persist=true
Nope. From crontab(5)
The RANDOM_DELAY variable allows delaying job startups by random amount
of minutes with upper limit specified by the variable. The random scal‐
ing factor is determined during the cron daemon startup so it remains
constant for the whole run time of the daemon.
I'll take payout from your lost bet now
https://dyn.manpages.debian.org/experimental/cron/crontab.5....?
OnCalendar=00/6
systemd-analyze calendar --iterations=6 '0/6:00:00'
https://www.freedesktop.org/software/systemd/man/latest/syst...
$ systemctl cat public-inbox-watch@.timer
# /etc/systemd/system/public-inbox-watch@.timer
[Unit]
Description=Periodic fetch of public mailing list
[Timer]
# twice a day
OnCalendar=*-*-* 5,17:35
RandomizedDelaySec=1h
Persistent=true
[Install]
WantedBy=multi-user.targetJust use Cron. It does one thing and does it well.
Mon,Fri *-01/2-01,03 *:30:45
[0] <https://www.freedesktop.org/software/systemd/man/latest/syst...>
[1] This question is sarcasm. SystemD is often like this... dead simple things look dead simple, but complex things are -if they're possible at all- at least as complex as they are everywhere else.
Cron syntax is simpler for the easy cases because cron tries to do less. It ignores years and seconds entirely, and doesn't try to adhere roughly to ISO8601 ordering and field separators, instead using space universally for field separation and euro-style least-to-most significant field ordering. I like ISO8601, so I get along with systemd's style better, despite it introducing slightly more cognitive load.
The only thing that threw me for a loop and seems like "special magic" was
> "Mon *-05~07/1" means "the last Monday in May."
But good luck doing that in one line in cron.
Some cron-style libraries seem to support L/W/# for last / nearest-weekday / nth of month, but I don't know if any system crons do. (cronie? dcron? I don't think so. fcron? bcron? I don't see it there either.) '#' is syntactic sugar for DOW + 7-day range, while L is covered by the above quoted syntax.
If your cron has that kind of syntax, then for a case like "weekday closest to 1st of month", "W" is more convenient than writing 3 systemd timer rules to cover the three cases (weekday day 1, monday day 2, friday last day of month), but that's a big if. Generally you'd have to write 3 rules in cron anyway.
I found this amusing when in combination with
> The only thing that threw me for a loop and seems like "special magic" was
but -regardless- a careful reader notes that I never said that the Timer scheduling syntax was illogical or poorly thought out. It's at least as complicated as crontab-style time syntax, which was my entire point.
Related: Not that it's ether part of the core scheduler syntax or necessarily as nice as having it in the core syntax, but my crontab(5) suggests that one can use things like date(1) to get more fine-grained control over the time of execution:
As noted above, skip values only operate within the time period they´re attached to.
For example, specifying "0/35" for the minute field of a crontab entry won´t cause
that entry to be executed every 35 minutes; instead, it will be executed twice every
hour, at 0 and 35 minutes past.
For more fine-grained control you can do something like this:
* * * * * if [ $(expr \( $(date +%s) / 60 \) % 58) = 0 ]; then echo this runs every 58 minutes; fi
0 * * * * if [ $(expr \( $(date +%s) / 3600 \) % 23) = 0 ]; then echo this runs every 23 hours on the hour; fi
Adjust as needed if your date(1) command does not accept "+%s" as the format string
specifier to output the current UNIX timestamp.
Two options:
0 0 25-31 5 1 0 0 L 5 1For cron, despite years of looking at it, I can never remember what those numbers mean and I need to Google almost every time.
Looking at the other examples on that page, I'm gonna say that it's only arguably easier to read for basic stuff... especially if you're familiar with the syntax. The complex stuff is -at best- just as difficult.
Eh sometimes, but you can get pretty far with one of two approaches:
1. Careful use of Requires= and Wants= to group your scripts into chains of jobs, which achieves fixed parallel (though at 100s of jobs, I hope you're generating those unit files with a tool like Puppet or https://github.com/karlicoss/dron or something and not doing this by hand).
2. Even better, just use a lockfile. `ExecStart="flock -F $TMPDIR/mylock <command>"` is pretty hard to beat. Use -F so as not to confuse KillMode and resource accounting and you're golden. Just don't use flock(1) timeouts; let systemd handle that. Heck, if you have that many cron jobs, you should be doing this even if you don't use systemd; otherwise job latency changes can cause reboot-style thundering herds out of the blue.
If you need semaphore behavior and still don't want a real job queue, waitlock (https://github.com/bigattichouse/waitlock) and many other CLIs have you covered.
2. If this for some reason fails, misconfiguration or unexpected shutdown, you could have a failure that's hard to track or debug
These are fine with a few services chained together, but this requires a shallow depth of dendencies. To have these theoretical hundreds of jobs chained together like this isn't practical or safe.
2) From the crontab(5) on my system:
The CRON_TZ variable specifies the time zone specific for the cron ta‐
ble. The user should enter a time according to the specified time zone
into the table. The time used for writing into a log file is taken from
the local time zone, where the daemon is running.
Systemd enables cleaner, simpler syntax for common cases. Instead of "59 23 * * *", simply "23:59". Instead of "0 0 * * sun[day]", simply "sun[day]". 1st of the month and don't care exactly what time? Instead of "0 0 01 * *", simply "*-*-01".
Systemd started with the principle that they wanted to accept ISO8601 timestamps, then extended that with lists, increments, ranges. They developed that into a superset of basic cron capabilities, while maintaining similar syntax as best they could for increments and lists; they diverged for ranges and nth-from-last because those conflicted with the - date separator. They repurposed the little-used ~ cron randomization operator in the process. (systemd uses a separate RandomizedDelaySec line for that, which is also arguably superior because it randomizes per trigger, not on initial load of the timer)
The alternative is how AWS does it. They have separate cron() and at() syntax. at() takes timestamps. cron() takes cron+quartz syntax, mangled to remove seconds. They also have rate(value units), because apparently cron syntax is too complicated for most people. I'm sure in theory they did it to support "every 7 minutes" cases. I bet if you surveyed actual rate() AWS schedules, 99% would be cases that evenly divide a larger time unit, and could therefore be implemented with cron().
Cron syntax is more of a mess than I thought. While basic system crons don't support any advanced features, AWS and Cloudflare have adopted LW# capabilities, derived from quartz job scheduler syntax (Cloudflare references quartz explicitly). Quartz has extra fields for seconds and year (year optional, so it must go last, which breaks d m Y sequencing) so its syntax has 6 or 7 fields vs the standard 5. CF uses 5 fields, no seconds or years. AWS uses 6 fields, years but not seconds. Whenever you interact with cron-derived systems, you have to remember not just whether they support quartz-style syntax, but whether they support seconds and/or years. Is that really simpler?
The one feature I found interesting, "W", may not work the way I anticipated according to quartz's documentation. I thought "nearest weekday to the 1st of the month" might be useful because it maps to some billing cycles, but quartz (at least through 2.3) couldn't do that. The 2.3 docs say that 1W, if the 1st falls on a Saturday, triggers on the 3rd because it won't jump backwards over a month boundary. That was my only interesting use case for "W". Systemd does everything else, though it requires more verbosity in some cases, it's simpler in many others. 2.4/2.5 are from the last couple of years, and their docs are restructured and don't mention that, so maybe it was finally fixed? I'm not about to install java crud to test.
I didn't find the *-*~01 syntax that strange (last day of every month). The "fri *-*~07/1" syntax for "last friday of every month" is what I found magical, but it makes sense now that I've thought about it more.
What would you prefer for 3rd friday of the month? Cron (5-field, quartz syntax)
0 0 ? * fri#3 (special syntax is needed because cron uses logical-or to combine day and weekday fields)
fri *-*-15..21 (works because systemd day and weekday are joined with logical-and)
[aws] https://docs.aws.amazon.com/scheduler/latest/UserGuide/sched...
I disagree. It's all a matter what you're accustomed to. On top of that, what's at least as important as clear syntax for simple cases is the establishing of a handful of rules that well serve both the simple and complex cases. Consistency that leads to complicated cases being made easy should not be avoided just to make the simple cases extremely welcoming to the newcomer. Note that I'm not saying that 'systemd-crond' falls into this trap, but I am saying that oh so many tools that aim to be "approachable by newcomers" enthusiastically throw themselves into it.
> Cron syntax is more of a mess than I thought.
Nah. It's just there are a very large number of cron implementations. If The Systemd Project were actually a thing that people could make an independent reimplementation of, we'd see at least as much inconsistency in 'systemd-cron' scheduling syntax as people extended it in their reimplementations to meet their needs.
> What would you prefer for 3rd friday of the month? Cron (5-field, quartz syntax)
I don't understand why the "day of month" column has a "?" instead of a "*", and I don't understand why "fri#3" isn't rendered as "Fri/3", but whatevz. I prefer that to the systemd syntax.
FWIW, I think you can directly translate your systemd scheduler line into this for most crons:
* * 15-21 * Fri
Note: The day of a command’s execution can be specified in the following two fields — ’day of month’, and ’day of week’. If both
fields are restricted (i.e., do not contain the "*" character), the command will be run when either field matches the current time.
For example,
"30 4 1,15 * 5" would cause a command to be run at 4:30 am on the 1st and 15th of each month, plus every Friday.
Everything else is simpler. You don't have to count fields to figure out whether you're dealing with a day vs a month, or an hour vs minute, no matter how the fields are aligned or how long or complex they are. You can scan left and right and look for space and dash and colon to figure out where you are. The formats for time and date are distinctive and ubiquitous. You don't have to guess what you're looking at when you just see one of them in isolation.
Very little follows from cron's format. You have to learn which fields are which for your particular implementation: does it include seconds? You have to learn the bizarre interaction between day and weekday, to do something non-trivial with them. You have to carefully keep track of which field you're in because you can't determine that from neighboring syntax. The visual clutter, if you want to call it that, of systemd's iso8601-based syntax only applies if you're doing complicated rules. In the vast majority of cases it would simply be an ISO8601 timestamp with various values replaced with *'s, and at most one /.
That first impression you had is what I thought too, before diving into it for my previous replies. But no, cron logic is that day and weekday are ORed. It lets you do clever things like run a script on the 1st and 15th of every month, AND on mondays, all in one line. I don't think I've ever seen such a thing in the wild. I'd go so far as to say it's stupid, because it's likely to be misinterpreted if anyone did use it that way. All the other fields are ANDed. I would say that systemd's AND logic is better... and more consistent, which is your own acceptance criterion.
At least busybox's cron implements it that way:
if (line->cl_Mins[ptm->tm_min]
&& line->cl_Hrs[ptm->tm_hour]
&& (line->cl_Days[ptm->tm_mday] || line->cl_Dow[ptm->tm_wday])
&& line->cl_Mons[ptm->tm_mon]
) {But first I‘d check the behavior of other crons ;-)
# Run a command on the last day of the month. Only starts checking at midnight towards the end of the month. Assumes GNU date, which is fair if we're discussing a Linux-only cron-alike.
0 0 28-31 * * if [ $(date +\%d) -eq $(date --date="$(date +\%m)/1 + 1 month - 1 day" +\%d) ]; then /usr/local/bin/runCommand.sh; fi
If I were doing this for real, I'd either switch to a more capable cron, or take a serious try at the date math and then wrap it up as a standalone helper. Or I guess I'd look to see if someone already built that helper. ...I guess...
It ultimately doesn't matter "for real", because almost nobody without some horrific legacy system to integrate with would need to use anything more than lists, ranges, and increments. There's a reason nobody's added these features to system crons. Clever trigger times for events that don't really need to be triggered at clever times... sure, but lacking that capability wouldn't change anything, the trigger would just be a more boring one.
Events like "last Friday of the month" or "nearest weekday to the 1st of the month" or even "Friday the 13th" are more for business logic, not system crontabs.
I am no cron scholar, nor am I young enough to have the energy required to do an exhaustive survey... so -sadly- I don't know of one. It seems like there'd be one being used internally in at least one business in the world, though, yanno? That just seems like the sort of thing that at least one bored programmer would take a few days to crank out.
> Events like "last Friday of the month" or "nearest weekday to the 1st of the month" or even "Friday the 13th" are more for business logic, not system crontabs.
Oh quite possibly, yeah. But, system crons are quite fine for one-shot things that are re-run on a regular schedule... as well as things that don't have complicated scheduling and/or retry requirements. If you're working on a Big Enterprise Project [0], then you're almost certainly going to have a scheduler inside of you, and -IME- you're very likely to use it to do a lot of that BEP's scheduled tasks. [1]
[0] Or a tiny one wearing the clothes of a BEP
[1] ...if for no other reason than it requires little effort to get the BEP's data and code into its internal scheduler, and can be a huge pain in the ass to get it into an external one.
Nope. I don't give a shit how "clean" something looks, [0] I just care how difficult it is to learn and either retain or remind yourself of the syntax when you need to do something complex with it. In that regard, [1] systemd-crond's scheduler syntax is at least as complicated as that of most crons'.
> You don't have to count fields to figure out whether you're dealing with a day vs a month, or an hour vs minute, no matter how the fields are aligned or how long or complex they are. You can scan left and right and look for space and dash and colon to figure out where you are.
So, you have to learn a tool-specific format. Just like you do for crontab. I've read both man pages, and I'm 100% unconvinced that systemd-crond's scheduler syntax is actually simpler. I do agree that it can express specific schedules that one cannot in the syntaxes used by most cron implementations.
> Very little follows from cron's format. You have to learn which fields are which for your particular implementation:
Sure. And I expect that exactly nothing other than systemd-crond uses its weirdo syntax. And -as I mentioned in the comment you're replying to-
Nah. It's just there are a very large number of cron implementations. If The Systemd Project were actually a thing that people could make an independent reimplementation of, we'd see at least as much inconsistency in 'systemd-cron' scheduling syntax as people extended it in their reimplementations to meet their needs.
On this I agree with you. Changing the rules for the Day Of The Week column (but only when it's used in combination with the day of the month column) is catastrophically idiotic. The only time I've ever had need to use the DoW column was to schedule something to run weekly... so I've been blissfully unaware of how incredibly idiotic that behavior is.
[0] If anything, I've noticed a strong inverse correlation between how "clean" a syntax is regarded to be and how easy it is to both do complicated things with it, and come back to those complicated pieces of work six+ months later and be able to understand what the hell you did.
[1] With the notable exception of cron's DoW column.... read on for further discussion.
Nobody is all that happy with cron's limitations. That's why there are so many variants, with varying features. When a variant has more than 5 fields, the syntax becomes ambiguous... not to a parser, but to whoever's reading and writing the lines, unless they already know the exact cron variant in use. Google, the one large cloud provider that seems to have stuck with basic cron syntax without quartz-scheduler extras, even invented their own pseudo-natural-language scheduling syntax to fill in for features cron lacks, or for people who don't like cron syntax.
Systemd's is definitely novel, but as I keep repeating, it's ISO8601 + cron. An alternative would be some extension to ISO8601 interval syntax, but that looks clunky and has probably never been used outside of niche applications for data interchange, probably inside xml files; it's used in https://docs.cibseven.org/manual/2.0/reference/bpmn20/events... (syntax at https://en.wikipedia.org/wiki/ISO_8601#Durations ). I think systemd's syntax is better than that.
The direct benefits, even if cron syntax suffices: Calendar timers are integrated with systemd, with all its process-management capabilities[1]. Unifying to systemd timers also avoids duplication of functionality ("is this scheduled process managed by systemd or cron?"). Cron doesn't let you `systemctl list-timers --all`. You'd have to grep through cron logs, possibly (and ironically) using journalctl, or have those logs already in a management dashboard, and that still doesn't let you see the next time the timer would fire. Cron also doesn't let you beta-test cron lines like you can with `systemd-analyze calendar --iterations 10 "Fri *~8..14/1"`
There's even a systemd-cron translator/generator that's in the official debian/ubuntu repos, and in Arch's AUR of course, which avoids the need for system cron but allows you to retain existing crontabs by importing them as systemd timers. They won't be quite as well integrated because it uses defaults for task names and other parameters, but the crontabs get run... without cron.
I'll repeat my opinion, as expressed in my first post:
To do nontrivial scheduling you'd use the entirely-obvious-and-intuitive syntax described at [0]. For example:
Mon,Fri *-01/2-01,03 *:30:45
Who'd ever want to go back crontab format for nontrivial scheduling? [1]
[0] <https://www.freedesktop.org/software/systemd/man/latest/systemd.time.html#Calendar%20Events>
[1] This question is sarcasm. SystemD is often like this... dead simple things look dead simple, but complex things are -if they're possible at all- at least as complex as they are everywhere else.You keep saying systemd calendar format is "at least as complicated", but it seems to me the converse is true. Simple cases are simpler to represent in systemd because fields are optional in various circumstances; complex cases, where they can be implemented in system cron at all, look equivalently complex. Removing the weekday constraint, rounding to minutes, and removing the star for year since it's optional:
31 * 01,03 01/2 *
01/2-01,03 *:31