Swâmi Petaramesh
2014-Aug-24 17:25 UTC
REPORT: Still a lot of BTRFS freezes in 3.16.1-1-ARCH
Hi there, This is to report that I'm still having quite systematic BRTFS freezes on an ArchLinux running latest 3.16.1-1-ARCH kernel. Interestingly enough, I have several latops with the exact same setup : Arch Linux with 3.16.1-1-ARCH kernel, fully running on BTRFS (with LZO compression) over LVM over LUKS. All machines have hourly BTRFS snapshots performed by SuSE « snapper » utility. The difference between the laptops only relate to the disk (HD vs. SSD) and CPU "power". I have : - Intel Core I3 with mechanical 1TB Samsung HD: NEVER hangs. - Intel Celeron CPU 1007U @ 1.50GHz with Crucial_512GB MX100 SSD: NEVER hangs. - Asus EeePC, Intel Atom CPU N450 @ 1.66GHz with SanDisk 120 GB SSD: VERY OFTEN HANGS, only since 3.15 kernel, and I'm pretty sure that the hardware itself is OK. It seems that the only machine on which BTRFS crashes is the one on which the CPU is weak & slow relatively to the SSD (so the I/O, considering compression and encryption, is quite surely CPU-bound). The typical situation causing a machine crash is : Performing a lot of updates using "pacman -Syu". The typical freeze happens while applying the updates (not while dowloading the packages). The typical result is : - The update process stalls. - The rest of the machine AT FIRST seems unaffected - After a while eventually everything that needs disk access freezes - A fierce reboot is finally needed. The typical outcome is : - After reboot, a number of files belonging to the packages that were in process of updating, or had been updated last, appear as having a size of ZERO bytes (so what depends on them is typicaly broken). (If the write to disk is ORDERED, this isn't supposed to happen, is it ??) - Forcibly reinstalling the packages for the broken files usually fixes things up. I've had this exact experience at least 5 times on this machine now, always exactly as described. I've never seen any such issue on my other machines on which CPU power is better relatively to the disk speed... Hope this can be of any help ? I'm longing for a kernel fix cause I fear something could seriously break in the end... Kind regards. -- Swâmi Petaramesh <swami@petaramesh.org> http://petaramesh.org PGP 9076E32E -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html