ZFS On Linux Is Called Stable & Production Ready

[phoronix]

Phoronix: ZFS On Linux Is Called Stable & Production Ready

ZFS on Linux is now considered production-ready by one of the biggest contributors to the ZFSOnLinux project...

http://www.phoronix.com/vr.php?view=MTc4NTM

[kernelOfTruth]

Phoronix: ZFS On Linux Is Called Stable & Production Ready

ZFS on Linux is now considered production-ready by one of the biggest contributors to the ZFSOnLinux project...

http://www.phoronix.com/vr.php?view=MTc4NTM

works for me as daily driver, backup media, laptop, etc.

endured dozens of hardlocks, lockups, resets without any data loss or corruption

[ryao]

I'll have some updated ZFS Linux benchmarks on Phoronix in the near future.

I hope that Michael will address the issues that were present in his previous ZoL benchmarks:

https://news.ycombinator.com/item?id=8292566

By the way, ZoL 0.6.3 supports Linux 2.6.26 to Linux 3.16 while ZoL HEAD adds support for the Linux 3.17 release candidates. I have confirmation from another ZoL contributor that HEAD works on 3.17_rc4.

[jvsalo]

RAID benchmarks?

I would love to see some RAID benchmarks for a change. I know Michael has stated he doesn't have adequate hardware, but still it looks like he's receiving thousands of $ worth of equipment every once in a while

For software RAID benchmarks, just a bunch of disks and perhaps a HBA (probably doesn't even need one) wouldn't be too steep.

It could also be interesting to compare performance of the send functionality in ZFS and btrfs.

[dungeon]

ZFS On Linux Is Called Stable & Production Ready

I thought i had one year of sleep after reading this .

2013 ? The first stable release of ZFS on Linux.

http://open-zfs.org

Or maybe every year we need to call it stable

[ryao]

I thought i had one year of sleep after reading this .

http://open-zfs.org

Or maybe every year we need to call it stable

The purpose of the blog post was not to claim that ZoL has just become production ready, but to provide an explanation why ZoL is production ready. There was a severe deficit of information on this subject prior to the blog post. It is intended to correct that.

[kernelOfTruth]

The purpose of the blog post was not to claim that ZoL has just become production ready, but to provide an explanation why ZoL is production ready. There was a severe deficit of information on this subject prior to the blog post. It is intended to correct that.

Hi Richard,

after reading through parts of your articles (not through all 4 yet)

can you shed some light how IOMMU is used in conjunction with ZFS ?

if I understood correctly IOMMU is the counterpart to intel's VT-d - so this has to be explicitly enabled in BIOS (if not on by default) ?

I suspect that

zcat /proc/config.gz | grep -i IOMMU
CONFIG_GART_IOMMU=y
CONFIG_CALGARY_IOMMU=y
CONFIG_CALGARY_IOMMU_ENABLED_BY_DEFAULT=y
CONFIG_IOMMU_HELPER=y
# CONFIG_IOMMU_SUPPORT is not set
# CONFIG_IOMMU_DEBUG is not set
# CONFIG_IOMMU_STRESS is not set

is not enough ?


what role does NCQ play in the guaranteed ordering of data ? does it have to be disabled ? (queue set to 1)

most if not all harddrives driven by S-ATA don't support flush/FUA (Write cache: enabled, read cache: enabled, doesn't support DPO or FUA) - so this means the writecache has to be disabled to guarantee the order and thus integrity of data ?


Many thanks in advance and all the best at your new job

[NotMine999]

I Run ZFS On Gentoo...And Like It

Phoronix: ZFS On Linux Is Called Stable & Production Ready

ZFS on Linux is now considered production-ready by one of the biggest contributors to the ZFSOnLinux project...

http://www.phoronix.com/vr.php?view=MTc4NTM

I have "spun" 100+ TB (raw capacity) of ZFS storage across 5 servers for "personal use" for almost a year now on almost 24x7 basis...and without any issues with ZFS.

All of these systems run Gentoo 64-bit code with the current "gentoo-sources". The hardware is Supermicro X10SBA boards with 16GB of RAM, separate boot drives, and WD 4TB RED HDDs in a RaidZ1 setup. All systems are on very solid "pro-sumer grade" UPS systems by Eaton. Network performance is, well, pretty astounding across a switched GigE network; uploads and downloads of large (>2GB) files can easily "max out" the GigE bandwidth with sustained transfer measurements showing in the 700-900 Mbps range.

Sure, 16GB is technically not supported on that board by the vendor; it will beep a few times during POST. Sure, it is not ECC memory; ZFS zealots would demand their ECC, but those boards cost more $$. Sure, that board might not be considered "data center grade" by some; Supermicro does build a solid product, seriously. Sure, I could probably consider running RaidZ2 if the information was "really critical". Sure, my experience might not be adequate proof to others that ZFS is "production ready"; no hard data to track ongoing performance and so on.

But...Hey There! It really does work, it works very well, and it runs incredibly solid & without user intervention except for "upgrades".

It is most unfortunate about the licensing of the source code due to Oracle. On the other hand it has made me learn how to handle updating "out of kernel" modules on Gentoo.

[ryao]

Hi Richard,

after reading through parts of your articles (not through all 4 yet)

can you shed some light how IOMMU is used in conjunction with ZFS ?

if I understood correctly IOMMU is the counterpart to intel's VT-d - so this has to be explicitly enabled in BIOS (if not on by default) ?

I suspect that



is not enough ?


what role does NCQ play in the guaranteed ordering of data ? does it have to be disabled ? (queue set to 1)

most if not all harddrives driven by S-ATA don't support flush/FUA (Write cache: enabled, read cache: enabled, doesn't support DPO or FUA) - so this means the writecache has to be disabled to guarantee the order and thus integrity of data ?


Many thanks in advance and all the best at your new job

The IOMMU is a technical term that AMD uses while VT-d is Intel's marketing term. It does need to be enabled in the BIOS. You can check dmesg to see if it is on. It will sometimes break poorly written device drivers.

As for the NCQ, there should be no reason to disable it. There exist block devices that reorder across flushes. If an AHCI or SCSI drive is one of them, its firmware is broken. It is typically RAID that will reorder across flushes, but that is not the only place where it can happen.

FUA is another animal. ZFS relies on flushes rather than FUA to ensure consistency. I am thinking of modifying ZIL to utilize it on slog devices, but drives that ignore FUA are a concern for that.

[ryao]

it has made me learn how to handle updating "out of kernel" modules on Gentoo.

genkernel should make this easy. You can even provide your own .config, which it will use if it does not have an existing one cached in /etc/kernels.