Related to this.
I'd like to take advantage of an OS switch to upgrade to BTRFS.
BTRFS claims to offer a lot (data-loss resiliency, self-healing if RAID, checksumming of metadata and data, compression, snapshots). But it's slow when used with
fsync-intensive programs such as
eatmydataand the crappy
apt-btrfs-snapshotprograms) and I won't setup a RAID :p.
EXT4 allow metadata check-summing only and doesn't compress data.
In 6 years, I had to reinstall my OS twice because of HDD corruption (after flight trips). The first making the laptop unbootable, the second bunch of corruptions was identified thanks to a corrupted film and then md5sum check of the OS binaries. (SMART tells me the disk is sane). The lappy currently behave quite strangely. I don't know if the hardware or the software is to blame but I suspect the hardware (it all began right after a flight, once again).
Would you advise to switch to BTRFS for a laptop because of data compression and check-summing or should I stick with EXT4?
(I don't care about which is "best" relative to whatever variable but I have almost no experience with BTRFS and would like some feedback)
Let's be clearer: BTRFS is still flagged as experimental, I know, but SUSE says it shouldn't anymore. So does Oracle (I know who Oracle is). And a bunch of distributions already propose BTRFS for installation and most of them are planning to switch to it in the next few months.
- Backups of corrupted data are worthless. I don't understand why I seem to be the only one to bother. Isn't that common sense? In the meanwhile:
- Stop telling me I should do backups: I already do.
- Stop implying backups are just enough to keep my data safe except if you are willing to give me TBs of free space to do years worth of backups.
- A corrupted file =/=> Linux complaining. So:
- Don't assume your system/data are sane just because the OS is booting.
- I hope you understand that I prefer (meta)data checksumming to an over-engineered and bloated piece of software that would inconveniently do half as a good job as BTRFS to check the data integrity.
Is that more clear now that I am not asking for which FS is "better"? The question is, given that I regularly do backups, is BTRFS still too experimental to be used for its data-integrity checking functions or should I stick to EXT4?