Another Look At The Bcachefs Performance on Linux 6.7

[NotMine999]

good luck with bitrotting data!

Are you just passing hot air here? Or do you have actual proof to back up your claims?

I have used Ext4 for all my boot volumes ('/' root & /boot) for the past 15 years simply because it is SOooo easy to recover if something fails.

Have I had SSDs fail? Yes I have. We just install a new SSD; load up a new copy of the OS (netboot that install so no shuffling of CDs or remembering that specific USB key that rarely gets updated); restore from backup any config files that are different from default; then reboot the system back to where it should be. Application restoral, if it resides on the 'System' drive (but it should not per "house rules"), is handled differently in my shop.

Sure, disk imaging using something like Clonezilla would be a useful idea for our 'System' drives, but my administrative/development/teaching time is limited. The few "house techs" on staff know how (and have documented procedures) to replace hardware, netboot an OS install, and restore config files from backup ... among other tasks.

YMMV

[edgmnt]

I just dont like sysvinit bash scripts although its more flexible, readable and configurable i just dont have enough experience with bash script to make a proper init script for the life of me.

It's not just you, non-systemd stuff tended to be duct-taped and essentially broken one way or another, if it could even be done properly. With systemd the executables don't even need to write pidfiles or daemonize themselves anymore. Less privileges, less race conditions and it just works with minimal configuration. If you really want you can set up stuff like socket activation and ensure the stuff comes up properly when and if needed.

[edgmnt]

Ext4 still looks competent though. Thanks for the benchmarks.

To be fair, it's not really hard for some of those benchmarks.

[curfew]

Ah, the denial is strong here. Pointless bickering about everything except the topic of the article. Bcachefs performed poorly and no-one is ready to acknowledge it.

[Snaipersky]

Ah, the denial is strong here. Pointless bickering about everything except the topic of the article. Bcachefs performed poorly and no-one is ready to acknowledge it.

The takeaway is that it performed significantly better than in initial testing, when it was committed to mainline with defaults building with every debug feature available. So while it isn't the fastest, it pulled a couple wins and offers a good set of features without cripplingly poor performance.

[leSh0uno5ee]

... the RAIDZ of Theseus.

this.... this is beautiful

[andyprough]

good luck with bitrotting data!

Cool story, leave a link to the bug reports you submitted, I'm sure the devs would have been all over this horrendous data destroying conspiracy you uncovered. We should all be able to see for ourselves how the devs addressed it.

[leSh0uno5ee]

Ah, the denial is strong here. Pointless bickering about everything except the topic of the article. Bcachefs performed poorly and no-one is ready to acknowledge it.

This is a good point.

I myself also have great expectations toward bcachefs so I'm happy seeing it lagging behind, because knowing Kent (not personally ofc, but his temper) this will be fixed soon just because of pure ego.

Again, as I've pointed out before, this was not a good test. This was kind of interesting, but not a good one. Why on Earth would anyone compare extX with btrfs, f2fs with bcachefs? And where's ZFS? Totally different FSes for totally different use cases. A good sysadmin/engineer/architect knows what kind of FS should be used for what purposes so this kind of test is pointless.

[cynic]

Are you just passing hot air here? Or do you have actual proof to back up your claims?

I have used Ext4 for all my boot volumes ('/' root & /boot) for the past 15 years simply because it is SOooo easy to recover if something fails.

Have I had SSDs fail? Yes I have. We just install a new SSD; load up a new copy of the OS (netboot that install so no shuffling of CDs or remembering that specific USB key that rarely gets updated); restore from backup any config files that are different from default; then reboot the system back to where it should be. Application restoral, if it resides on the 'System' drive (but it should not per "house rules"), is handled differently in my shop.

Sure, disk imaging using something like Clonezilla would be a useful idea for our 'System' drives, but my administrative/development/teaching time is limited. The few "house techs" on staff know how (and have documented procedures) to replace hardware, netboot an OS install, and restore config files from backup ... among other tasks.

YMMV

you would have save yourself a lot of time just googling what bitrot is instead of writing all this stuff.
bitrot is a very known phenomenon, no need to backup my claims here: just learn what it is and why a non full-checksumming filesystem as ext4 is vulnerable.

also, you risk backupping corrupted data and then restoring corrupted data.

[cynic]

Cool story, leave a link to the bug reports you submitted, I'm sure the devs would have been all over this horrendous data destroying conspiracy you uncovered. We should all be able to see for ourselves how the devs addressed it.

bitrotting is not a bug of the filesystem.
next time, instead of showing off and call other people conspiracy theorist, just use a search engine to at least undertand what you are talking about.