https://github.com/anatol/booster
Does this give any real world value for boot times or anything else?
I have no possibility to test this in VM so that’s why I’m asking if anybody has actually tried this and found benefits.
Removed by mod
I use booster and it’s cool. I don’t see any noticeable difference in boot times but the image generation is much faster. mkinitcpio would take several seconds while booster takes about one.
First time I tried it it didn’t boot because of something missing in the generated image. I tried a universal booster image (set
universal: True
in/etc/booster.yaml
) and it worked. Technically this builds a larger image than necessary but it’s still only 34MB and takes a second to build, so I never bothered to troubleshoot what was missing. The universal image even handles luks encrypted root partitions without additional configuration of booster (you still have to configure kernel parameters).Another issue I noticed is that if you use
grub-mkconfig
and your only initramfs is booster, it will generate an incorrect main boot entry. It will add booster as an option in “advanced options” so your system is still bootable if this happens to you. The quick fix is to manually add theinitrd
entry under the main menuentry ingrub.cfg
.mkinitcpio takes around 2-3 minites to run on my machine, booster should be faster right?
What’s your machine?
some old crappy asus laptop from 2012-2014 with no hyperthreading, slow ddr3 ram, dead gpu and a mechanical hard drive
Thank you!
boot time difference feels like in the realm of margin of error
the biggest difference however is that booster builds the initramfs much much faster while mkinitcpio slows down every kernel upgrade espcially on slower laptop cpusI’m wondering how both mkinitcpio alternatives work on non systemd boxes with full disk encryption. With both, I refer to dracut and booster. On its origins I believe dracut was pretty tight involved with systemd, and booster is developed/maintained by an arch developer/user if not mistaken, and arch supposes systemd, though none of those things actually mean non systemd boxes are not supported.
I’m also wondering if the initrds generated can be launched by grub (I do /boot partition encryption/decryption with grub), and I also do / full partition encryption with luks. This booster issue sort of indicates as of now booster initrd images can’t be loaded by grub…
Grub can load booster images, the issue is about incorrect grub.cfg generation.
What they’re saying in the issue is that
grub-mkconfig
will not create a correct “Arch Linux” menu entry for booster, but if you go to “Advanced options” and choose the “booster” menu entry it works. I can confirm this. It happened on the system I’m currently using.Specifically, the problem is that
grub-mkconfig
does not add the booster image to theinitrd
of the default menu entry. You can add it manually. For example I had to change thisinitrd /intel-ucode.img
to this
initrd /intel-ucode.img /booster-linux-zen.img
If I recall correctly this issue was not present last time I set up a system with booster. It might be a regression or maybe it only happens in specific system configurations.
understood, so no big issue with grub, cool !
I don’t have any input to add to the conversation, but I definitely would be interested to hear from anyone with experience using booster.
Well, I’ve never had any problems with mkinitcpio, so I’ve never tried booster or dracut
It generates initramfs images faster. Also the generated images are much smaller (30-40MB for fallback images).
The only problem is that the tpm2 unlocking doesn’t work currently.
I have tried dracut but not booster.
How was it? Where exactly you saw difference?
There might be some speed improvement (not that noticeable). But I can control the modules more freely than mkinitcpio, so I like it more.