Bonum Certa Men Certa

Messing With ZRAM Again Because IBM Software is Barely Possible to Document and Changes if You Do

Reprinted with permission from Ryan Farmer.

It turns out that when I was setting up Debian, I set up my “zram-generator” systemd configuration file wrong.



systemd is Hell because they never commit themselves to the idea that once you learn how to do anything with it, it will stay working, most of its components are full of bugs and security vulnerabilities, and many things replace something that some other part of the operating system was already doing a lot better.



I cringe every time I give it some new responsibility over my system, waiting for how it will go wrong. I would be using systemd-oomd to handle potential out-of-memory situations, but I’ve read so much about how it handles memory pressure bizarrely and worse than the kernel’s oom-killer, even with the Fedora defaults, that I figure I’ll just leave it alone unless it becomes “mandatory” at some point.



I hear (on Reddit) that systemd-oomd does crazy shit, like, “There are 32 GB of RAM in this system. 12 GB are in use. 20 GB are free. Let’s go on a murder spree and shut down some Chrome tabs and LibreOffice with unsaved work!”



ZRam is a compressed block device that you can use for a compressed swap file in memory. The idea isn’t a bad one, but IBM has made figuring out how to set it up unnecessarily painful because of course it is handled by systemd. You’d think setting up ZRam would be too simple for systemd to bring too much of its usual incompetence to, but nooooo.



Today I actually looked at zramctl and it told me the compression type was lzo-rle, not zstd as I wanted (and was the default on openSUSE and I’m sure that’s what it said when I set it up on Debian and started the service).



Whatever.



Looking around the Web, I found that I needed the line compression-algorithm = zstd in my /etc/systemd/zram-generator.conf file.



So I added it and rebooted, and checked sudo zramctl again and I had zstd compression. Yay!



But when I was looking at the manpage for “zram-generator”, it said that the method by which to specify the fraction of the size of RAM to use was obsolete.



zram-fraction = 1.00 does work, but it’s “obsolete”.



So now it tells me that the “current” way to do the same thing is zram-size=ram/x, where ram is the amount of ram and x is the amount to divide by.



So ram/2 would make the device half of RAM, ram/1 would make it all of your RAM, which is what I wanted.



So now my file looks like this:



# This config file enables a /dev/zram0 swap device with the following
# properties:
# * size: 50% of available RAM or 4GiB, whichever is less
# * compression-algorithm: kernel default
#
# This device’s properties can be modified by adding options under the
# `[zram0]` section, or disabled by removing the section header.
# Additional zram devices can be created by appending new `[zramX]`
# sections and setting the appropriate options for each device.
#
# See /usr/share/doc/systemd-zram-generator/zram-generator.conf.example
# and/or zram-generator.conf(5) for a list of available options.
[zram0]
zram-size=ram/1
compression-algorithm = zstd



I’m not exactly sure why IBM Red Hat keeps screwing around changing the way you do this. They make you memorize something only to make pointless changes to the way it’s done, but then the stupid thing doesn’t want to break existing setups, so they just say the old one is “obsolete”.



$ sudo zramctl
NAME ALGORITHM DISKSIZE DATA COMPR TOTAL STREAMS MOUNTPOINT
/dev/zram0 zstd 15.4G 4K 64B 4K 8 [SWAP]



At least this thing is finally set up right, I think.



They never let you get too sure that you did it right or it wouldn’t be “modern”.



As to the debate about zstd vs lzo-rle, they’re both decent choices.



I can see the logic of defaulting to either one, depending on who you think your users are.



If you have a multi-core x86 PC with lots of CPU performance to burn, zstd makes more sense because of higher compression ratios.



If you have some little ARM system that you bought down at the Micro-Center and don’t want to overload it, but still do want to use ZRam, lzo-rle is a respectable choice.



If anything, ZRam with zstd has gotten faster since the kernel developers have last updated the default and I suspect it deserves another look.

Recent Techrights' Posts

Corporate Media: Blame the People Who Enter the Abandoned IBM Buildings, Not IBM for Abandoning Workers in Pursuit of IT Sweatshops
When the media spreads falsehoods stocks can go up (a lot higher), but at whose expense and how long for?
SUEPO Munich Report on the Recent EPO Demonstration and Rolling Strikes That Continue to Grow
"increasing registrations for the 'rolling strikes' running until autumn"
Gemini Links 11/07/2026: Old Computer challenge, Poems, Antenna, and More
Links for the day
 
Blogs May be Making a Comeback (They're Not Fediverse, They Are Joined by RSS Feeds)
Don't fake expansion where none existed
ChromeOS and GNU/Linux in the United Kingdom Reach 11%
the UK shows signs of digital maturity
Canonical is Selling Microsoft, It Pays The Register MS to Sell Microsoft
It's all about money to them. And they call this journalism.
When Red Hat's HR Becomes the Same as IBM's HR (Bluewashing)
Red Hat keeps sacking very experienced engineers and adding temporary interns
GNU/Linux Growing in East Asia
Assuming this is more or less accurate, we could use a plausible explanation
Over a Week After Microsoft Discontinued Some XBox Models It Apparently Exits Some Markets Altogether
We seem to be witnessing the end of XBox
Links 11/07/2026: "Trademark wars of Influencer Culture", Xinuos Uses Copyrights Versus UNIX
Links for the day
North America: GNU/Linux Measured at 10%
To better understand what contributes to the gains
Following Corrections and Adjustments statCounter Sees GNU/Linux at 7.1%, an All-Time High
There is a lot of layoffs at Microsoft this month
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Friday, July 10, 2026
IRC logs for Friday, July 10, 2026
Links 11/07/2026: Wednesday-Saturday News Catch-up
Links for the day
Prioritising High-Importance News
In order to fully catch up with news we'll not publish many new articles until next week
The Register MS: "AI" More Than 80 Times in One Article. But It's Not an Article, It's Sponsored Keyword-stuffed Page.
The Register MS is being paid to actively promoted this scheme
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Thursday, July 09, 2026
IRC logs for Thursday, July 09, 2026
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Wednesday, July 08, 2026
IRC logs for Wednesday, July 08, 2026