In other words in the latter case of a flipped bit getting through to the filesystem, how do you know you are detecting the flipped bit at all, and are you So it seems to me that my claim still stands that the ZDnet artikle is way overblown. At work, most of my peers and managers are often satisfied to know that someone on the staff understands this stuff so they can work on other things instead... Unless one is a hard drive manufacturer, OEM licensee, or reasonably-talented hacker with the right equipment & software to access the drive firmware, claims about actual in-the-field URE rate knowledge are his comment is here
But putting all that aside, I'm not sure what you are hoping the manufacturers should do. With 6TB drives I am beyond the math. I found this article but it's just one. I'm very curious?
Note that we are using TB and PB, not TiB and PiB. 1TiB is what Windows would report as a TB and is 1,099,511,627,776 bytes. 1TB is what drive manufacturers call in day to day operation, you probably read well over 11.3tb of data off your hard dive in a year and we don't seem to have this mass drive failure or Bad blocks, sectors , areas on the harddrive, that the drive itself and the OS is not aware of. More efficiency-minded vendors will instead pack the data on a 4kn drive with multiple 512e writes, a CRC for each 512e write, plus a CRC for the 4kn sector as well.
ZFS users variable block sizes and applies parity to those blocks. If a customer insists on RAID5, I tell them they can hire someone else, and I am prepared to walk away.I haven't even touched on the ridiculous cases where it takes I don't mean to offend, but it seems that you basically just really, really want single disk parity to be OK, and that's about all there really is to that. Zfs Ure Probability of a read error while reading all of a 100GB volume using SATA drives 100*(1-(1-1/(1E14))^(100E9*8)) = 0.80% (rounded off) So we’re getting about the same answer using bits instead of
Ideally firmware and OS work together hand in hand.Very often users or the operating system won't even know that a "URE" is present. That 1014 number is a worst-case scenario. Your array has failed. You can download Solaris here: http://www.oracle.com/technetwork/server-storage/solaris11/downloads/index.html Try out the 11.3 beta; there are many great innovations and improvements.
While his description may adequately explain the behavior of some drives, I have a couple more observations that might be useful to explain your data. Ure Rate It's interesting to me how closely related the problems of adding new disks to a pool and setting up a pool with differently-sized vdevs are. In later kernels, a read-error will instead cause md to attempt a recovery by overwriting the bad block. A spec of one non-recoverable error per 10^14 bits read would equate to one error every 12.5 terabytes read.
Interesting topic - I would also be interested if anyone has any specifics on the QC curves, rejection rates, causes of failure or manufacturing drive stats. Enterprise SAS drives are typically rated 1 URE in 10^15, so you improve your chances ten-fold. Unrecoverable Read Error Ure In that case, Drive 1 is still readable and will be used in resilvering the new drive, but the premature replacement means ZFS is trying to reconstruct the data from parity, Ure Raid The assumption is that the RAID controller will be able to recreate the unreadable sector in memory using the data found on the other drives in the RAID array - it
Putting this into rather brutal context, consider the data sheet for the 8TB Archive Drive from Seagate. this content I located two documents that might be of interest to you that are about the TLER feature - the first document seems to do a better job of showing how this Looking at metaslab.c in both the Solaris code and OpenZFS code, the teams took two different approaches toward solving a similar problem. If an internal link led you here, you may wish to change the link to point directly to the intended article. Unrecoverable Read Error Rate Ssd
For example, UREs tend to cluster together, which can be really unfortunate. So on a single 4tb hard drive, you would simply need to fill that drive and read the info back off it 4 times and it will give you an error. From his article: (1 — 1/(2.4 x 1010)) ^ (2.0 x 108) = 99.2%. http://crimsonskysoftware.com/unrecoverable-read/unrecoverable-bit-error-rate.html The friendliest, high quality science and math community on the planet!
And the risk is eight times larger on 4kn drives than on 512n drives. Raid 6 Sorry to be so verbose; I'm just excited a community exists where I can let loose my inner nerd a bit! I really enjoy talking about how ZFS and related hardware/software work.
Does that sound right to you? The most basic issue right now is that write heads on drives are more or less at their room-temperature minimum size for the electromagnet to change the polarity of a single These built-in data recovery techniques often work very well, by the way; while they are proprietary by vendor, techniques like reading the polarity of neighboring bits, off-axis reads, and more can Raid 5 There is a whole other class of problems where the failure is of a larger scope, the head crashed into the platter, contamination from outside is wreaking havoc, the manufacturing process
For example, this Seagate data sheet shows for Barracuda SATA drives the number for “Nonrecoverable Read Errors per Bits read, Max” is 1014 This is the same number the author used Checksum errors on read are extraordinarily common, MUCH more so on some drive vendor models than others. Dismiss Notice Dismiss Notice Join Physics Forums Today! check over here I am perplexed then at what the issue really is ?
I am disappointed that Darren appears to either not have a proper grasp of these (to me, simple) concepts, or glosses over them in a way which appears dangerous and misleading. permalinkembedsaveparentgive gold[–]txgsync 0 points1 point2 points 1 year ago(0 children) How long does it take for the ZFS versions to move downstream to say Solaris and OpenIndiana type variants? Log in or Sign up here!) Show Ignored Content Know someone interested in this topic? Once two drives failed, assuming he is using enterprise drives...there is a 33% chance the entire array fails if he tries to rebuild.
Another alternative is to segregate your data manually. For the sake of argument, lets say it is 5 times as expensive. It's a risk that is an issue at large scales. Below is a write-up that I posted somewhere that quotes the Linux documentation about how the Linux software RAID handles unrecoverable rear errors:----Synology NAS devices use software RAID, Linux md software
The HDD mechatronics and the SSD physics are complex and hard to get right in all cases and that’s where the URE spec comes from, these random failures to read data In kernels prior to about 2.6.15, a read error would cause the same effect as a write error. Thursday, May 21, 2015 Unrecoverable read errors Trevor Pott has a post at The Register entitled Flash banishes the spectre of the unrecoverable data error in which he points out that If a disk throws an IO error, that gets counted under the "read" or "write" error columns in the pool status.
Those of us with real data are typically so tightly gagged by confidentiality agreements in order to get that data that we can't really say anything specific about it... The zpool is rebuilding data from parity, and one sector of parity data was the victim of a bit of debris in the drive. With RAID 6 you'd have a second parity, usually diagonal parity, which you can then use to recover following a lost disk and a read error. This behavior is not at all representative of striping, but of a space-map(metaslab)-based allocation algorithm.