Btrfs Seems To Finally Have Failed Me On A Production System

[NateHubbard]

Michael, if you continue loosing money you should eventually stop.

This probably ins' t a good business and I don't see this changing in the future.
Make the entire website pro, if people think yours is a good service they will eventuslly pay. Otherwise amen.

You know, I like this site and have read it for some time. Coworkers and I often discuss various news issues they Michael posts. But, it wouldn't bother me in the slightest if he gave up on benchmarking entirely. I don't care about benchmarking really, unless it's something really specific, like btrfs vs. ext4 or whatever, but as for how different video cards and processors perform in games, I don't care at all.

If Michael turned this into a tech news site only, I'd be fine with that, and maybe he'd profit, idk.

[mike4]

NO I disagree. Hardware tests are what keep me to this site. The're lots of other Linux News sites, without tests, that I never visit at all.
So BTRFS failed because of the GIT Kernel? Hence no bug report needed, and if rather to kernel guys?

[Slartifartblast]

Michael

I told you a few weeks ago the numbers coming out of that i5-520M were crap compared to even an i3-370M and that there was probably an overheat issue and down clocking. No use crying over spilt milk now

There is something seriously wrong with those benchmark figures, look at the results for I5-520M and I3-370M. Are you telling me that an I5-520M is half as slow as an i3-370m ?

[nanonyme]

NO I disagree. Hardware tests are what keep me to this site. The're lots of other Linux News sites, without tests, that I never visit at all.
So BTRFS failed because of the GIT Kernel? Hence no bug report needed, and if rather to kernel guys?

Maybe benchmarking could be made a premium user only feature?

[ormaaj]

The entire point of rc kernels is testing. If you're not interested in reporting useful bugs then there's no reason to waste your time and money testing rc kernels. Maybe someone can deduce whether problem results from a bug based on those screenshots. Doubtful.

[TiberiusDuval]

SSD:s have nasty style of silently failing, hdd's at least usually have some symptoms telling user it is better to back up now.
Btw. considering NTFS and its resilience. I had once faulty system drive (never never again Seagates for my computer), dual booting configuration. Ubuntu 12.04 with ext4, and Windows Vista then. Ubuntu died out as soon as drive failed. Windows did boot for a while, though slowly and gave out quite litany of errors. But I managed to salvage files on Windows Users directory. After some reboots Windows finally gave up the ghost too. I was glad that I had most of my Windows files on second hdd, and also Ubuntu's home directory on that working hdd. No major data loss, just reinstalling both OS'es on new drive.

[AndyChow]

I'm not surprised. BTRFS gets corrupt all the time. Especially if you balance or defrag using compression, on a raid system. You can be almost sure it's not going to finish. I opened two different questions on StackOverflow (or a sister site) that got me high rep, but no solutions.

You can tell how stable BTRFS is based on how many times it appears in each changelog of a kernel release. "Fixes corruption problems that ..., Fixes issue from...".

BTRFS claims that it's so much more flexible than ZFS, you can add devices and remove others on the fly, and so on. But try it! Try to remove a drive on a 5 drive raid-1 that has lzo compression! I dare you! Four days later, it still won't be finished. And it a disk breaks, try to mount in degraded mode and rebuild the array. I dare you again! It will eventually start spitting unrecoverable errors left and right. Wait, don't I have redundancy? Isn't that the point? Oh, the metadata was redundant, and the data was supposed to be redundant, but you didn't balance recently enough, and even if you mounted with data redundancy, it doesn't always work so well. Sorry. We told you to have another backup.

Yeah, BTRFS is kind of a joke right now.

[A Laggy Grunt]

Two things:
1: For money, charge people to see more than one or two stats. A while ago, there was that article about NVidia using your openbenchmarking.org data very heavily. If they (and other for-profit hardware companies) are going to use it instead of doing their own testing, they need to be paying for it so you can keep doing this.
2: Please do a proper bug report or at least send the Btrfs people an image of that filesystem corpse. If they figure out how your filesystem got corrupted, they can fix it so this is the last time you have to wipe and reinstall, at least for that bug.

[ryao]

Writing an article is _not_ a bug report. Raise one properly and help get _your_ problem sorted

Last year, a mainline kernel filesystem subsystem maintainer suggested to me in private that Michael was annoying all filesystem developers equally. I will not name him for saying something to me in private, but I am pleased to see that is the case.

That said, both the btrfs and zfs developers hang out in IRC. I am sure that the developers in #btrfs on freenode would be happy to get the opportunity to solicit more information through IRC. It is better than a bug report by news post.

[FireBurn]

The problem is if the bug gets fixed in a timely fashion then Michael can't write more sensationalist pieces about btrfs not working