Bitrot can happen on ssds also. Only a file system that does some validation of the data can help you in case a drive reports bad data without signaling it has an error.
1 Like
Dell sas6 should be super cheap but has a 2tb drive limit. Good for zfs or btrfs or md devices . not a raid controller just an hba
Thank you for that, now i think i should probably go and backup all my family pictures and important documents on optical disks and make frequent backups.
Hi, I also really like to hear about ReFS.
I have 3x raid 1 in my 2 pcs. Apart from going with Linux, I'd really like if ReFS was good enough and trusty for what I need, while getting more familiar with Linux first.
While Raid 1 has been ok for me in the past, recently an old 2TB Samsung HD204UI developed an error without me noticing it. Luckily, I barely wrote anything to the drive, mostly old data. When I was trying to make a full backup, I ran into read errors of the data written just before that. I tried some tools like HDSentinel surface test / read-write-read, but drive's done for. I didn't luckily lose much data at all, but I definitely could see that Raid 1 is not the last word either.
Would ReFS solve this problem in the future? I do not need compression or boot capability. I just need something like Raid 1, but more reliable? What about Windows Storage Spaces, how reliable is that?
Thank you!
Thank you
That's the really insidious thing. If bitrot creeps in, you don't know unless you look at the files every time you back up. Unless you keep the backups forever and ever you couldn't recover after overwriting your good backup with a bitrotted backup. And your backup software doesnt report that this file you haven't touched in years just mysteriously changed. zfs and btrfs do that. Though btrfs doesnt seem to have a patrol read function.. I think they expect you to rsync that stuff offsite or otherwise access the files every so often so btrfs can do its thing.
+1
I think btrfs is more meant as a first line file system, aimed at functionality and performance above data security. ZFS is meant to be guarding data integrity. In the development stages of btrfs, RPM distros used to default to btrfs on every workstation install, but now, they don't do that any more, they default to btrfs for the system partition, and to ZFS for the /home partition, which in my opinion makes the most sense.
What distro is it that uses ZFS by default?
Edit: Oh I realize now you probably mean XFS.
Yup, sorry, lack of concentration. Fact is, btrfs doesn't have the same performance in user data applications yet, hence why they switched I suppose. Great with XFS is that it's also possible to deposit the journal on an SSD for instance. That is the lowest cost data integrity solution for workstations in my opinion, that doesn't at all affect performance. The use of SSD's with XFS works great for enhancing performance.
I don't exactly work in enterprise, but I manage IT stuff at a pretty small business. I certainly learned a good amount, so thank you @wendell!
The previous IT person here had a serious hard-on for RAID5 arrays, and we have had a few instances of bitrot. I always thought it was due to having vastly different versions of Office around (anywhere from 2003 to 2012) causing some files to explode, but there's a good chance it's bitrot -- I just never knew.
Looks like I have my work cut out for me, we have a couple machines with RAID5 arrays, one with 4 drives, the other with 6.
Question though! How bad, in your opinion, is RAID10? Our NAS is currently running 4x2TB drives in that configuration. I like the performance personally, but it always felt more volatile than anything. If it is, that might be the first place I go to start fixing stuff.
@wendell I work at a company of 100 people and I am sole IT person. We currently have a Dell SAN PS4100E if I remember correctly. I find It much more complex to use a SAN then using something like local storage or VSAN from vmware. All of my VM's, 23 of them run on the SAN, I am limited by 2 1gbe connections to the SAN so there is nothing more I want to do then get rid of the SAN. The company bought it back in 2013 so I am suck for a while, SAN is always limited network connection, and switch that is why I prefer DAS. You also add multiple points of failure with a SAN, your switch can die, your san can die. If my switch or SAN goes down, everything goes down vs half my VM's if I use local or DAS. Hardware failure will occur, why would I double my chance of that?
I am also a windows environment mainly. Only have 5 VM's on linux.
I am not sure if that san supports the anti-bitrot stuff. Would be a question of what OS it uses and what the underlying file system is that your stuff is stored in.
DAS is dangerous if only because DAS doesn't have antibitrot stuff either.
Even FreeNAS (TrueNAS) could do a redundant mirror. The last EMC box off the line before the dell buyout, much to my horror, ran windows XP embedded. No antibitrot stuff.
Normally with a san, you would have redundant multi switch connections. Two switches, two (or more) links, two drive cabinets, etc. Or at least a cabinet with multi-path I/O. iSCSI makes multipath easy so that it is actually much more redundant than DAS in practice. If your SAN has more points of failure than you would otherwise have, then you don't need a SAN. But for a lot of VMs and generic VM hosts, having a highly available redundant storage pool is not a super tall order.
Heck I did some redundant mirrored storage pool stuff with two ancient PowerEdge 2950s. Each of those had dual 1gb and it was fully redundant against switch failure, machine failure, drive failure, etc. two 6 drive raid 5 arrays. Throughput would saturate both 1gb links. And iSCSI multipath was almost point and click
Huh, wow its amazing(ly scary) how easy it can be to get bitrot and corruption these days. I think on my next build I'm gonna try to do stuff to avoid that, like sas and well pretty much everything you mention.
Curious here too.
I would prefer to use Hardware RAID + a proper data integrity and backup plan... and sticking to windows for now.
What did I learn?
A whole lot. These videos are great Wendell!
@wendell I didn't understand 70% of what was going on here. Please don't kill me.
So if i want to have 2 SSDs in RAID 0, I need to buy the LSI Raid controller and not use sotware Raid.
But if it's in RAID 0 then I wont be able to use TRIM or will I?
If you are using Raid 0 then this topic isn't really for you. Raid 0 has nothing to do with protecting your data. In fact Raid 0 increases your chances of data loss over a single drive as if either drive fails, all is lost.
That's not answering my question. Besides I already know that ^^.