Post AsejPvgXfbK19EYiMC by splitshockvirus@mstdn.starnix.network
(DIR) More posts by splitshockvirus@mstdn.starnix.network
(DIR) Post #AseT4vRNmkyB2eQm4O by tomcat@mstdn.starnix.network
2025-04-01T15:50:28Z
0 likes, 0 repeats
In a 5x10TB ZFS array, which would you use?Why?
(DIR) Post #AsecCtKaPbOgP37pBo by splitshockvirus@mstdn.starnix.network
2025-04-01T17:32:44Z
0 likes, 0 repeats
@tomcat Are these Hard drives or solid state?
(DIR) Post #AsehXDVhax98YBpsqu by tomcat@mstdn.starnix.network
2025-04-01T18:32:26Z
0 likes, 0 repeats
@splitshockvirusHDDs
(DIR) Post #AsejPvgXfbK19EYiMC by splitshockvirus@mstdn.starnix.network
2025-04-01T18:53:31Z
0 likes, 0 repeats
@tomcat If the drives are larger than 8TBs, Raid 6.ReasonWhen you need to resilver an Array you need to read all the data on said array in other words all the sectors on all the drives, in order to rebuild the replaced drive. During this time the chances that another sector is damaged or another drive fail are increased because likely all the other drives are the same age and the probability of another drive failing or losing a sector is significant.That's the rule of thumb reason.The more scientific reason but condensed down because I don't have the white papers on hand is that as drive data density has increased, drive reliability has not kept up with the same pace it has stayed relatively constant. Somewhere in the 8TB range of number of sectors read you'll get a bit that is misread or some type of transmission error, normally this isn't an issue because you have parity drives go check every sequence but when you are rebuilding an array you are unable to do that because your check no longer exists. When you have Raid6 or RaidZ while you are resilvering your array you are also performing a checksum validation on the data that is then being used to resilver the array, because you have enough for a parity calculation.There are more technical ways to describe what I just said but again I don't have them at hand.With SSDs I don't believe the same probability of read error exists.
(DIR) Post #Asel2toSfHEsaaPQo4 by splitshockvirus@mstdn.starnix.network
2025-04-01T19:11:47Z
0 likes, 0 repeats
@tomcat The word I was looking for was Unrecoverable Read Error (URE), which is the chances that your drive will not be able to read a bit.Drive manufacturers typically quote a URE to occur between 10-14 and 10-15 for fancy enterprise drives.That's 12.5TBs and 125TBs respectfully, SSDs have a URE of 10-16 which is 1.25PB.So not 8TBs like I said but still the chances of it occurring are significant when you are resilvering.
(DIR) Post #Asel5CkrD6K9uop4Xg by tomcat@mstdn.starnix.network
2025-04-01T19:12:13Z
0 likes, 0 repeats
@splitshockvirusThanks for your detailed explanation! I was leaning toward 6 because this is pretty much what I've been reading, but the extra capacity and lower overhead for writes is tempting.Still, this makes good sense. Once this month's expenses are paid, and I get money back for the PC I just built for my Dad (I'm not making any money, but fronted expenses for all the parts and built it for him), I'm pulling the trigger on a TrueNAS.
(DIR) Post #AselQAWyfpev6EKLOS by splitshockvirus@mstdn.starnix.network
2025-04-01T19:15:59Z
0 likes, 0 repeats
@tomcat TrueNAS is great!
(DIR) Post #AsemFAgXrM7KLZ7dIm by 10leej@mstdn.starnix.network
2025-04-01T19:24:59Z
0 likes, 0 repeats
@tomcat Use btrfs but in all honest raidz6 if you plan to add more disks otherwise raidz5 is perfectly fine for a 5 disk array.
(DIR) Post #Asf5oxpR2HQs1j3qfg by tomcat@mstdn.starnix.network
2025-04-01T23:04:36Z
0 likes, 0 repeats
@10leejThanks! The extra capacity is enticing