All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
ihostart again.... 2 of 3 storage nodes TOAST
Calin-iHostART.Com — Today at 12:42 PM
Hello @everyone we have sadly news for STOR-2 and STOR-3 , in the last few days several HDDs started to break down, adding new HDDs and until they resynchronized with the backup from the raid was ineffective, because others ended up breaking down at the same time so cause a data lost, so that this kind of thing doesn't happen again of problems in the future, I switched from raid 6 to raid 60 , from 2x redundancy HDDs to 4x redundancy HDDs , at the same time, for this data loss, all customers on these nodes will be offered 1 month free, the billing period will be updated in the following days those affected on STOR-2 and STOR-3 will receive an email on this tonightm soon as possible, with the tutorial to re-mount the /storage partition
At the same time, after we move to colocation ,we are trying to get more money to buy new HDDs from the factory, for stop usage refurbished HDDs anymore.
We are sorry for the problems that have arisen
Regards,
Calin
Comments
How is adding raid 0 to the configuration is going to add any kind of redundancy? Also refurbished HDDs should be perfectly fine as long as smart values are in order. It's usually not unpredictable if a HDD is going to break down soon.
Multiple RAID6 arrays.
ie. if he had 12 drive RAID6 before, he would configure 2x 6 drive RAID6 and do RAID0 over them, moving from 2 drive redundancy to 4 drive.
Hmm, yeah, kind of but the raid 0 is really just concatenating both arrays with a given piece of data being stored on one or other with actual redundancy still being as it were before or am i making a mistake somewhere?
If you take 12x Drive RAID6, which has 2 drive redundancy, split it to 2x 6x RAID6 you have 4 drive redundancy. the RAID0 is just to join their performance together, thus creating RAID6+0.
You have now effectively doubled up your redundancy, while still having one big volume, and in terms of read performance almost as much performance.
But if there is no mirroring in raid 0 how does it double redundancy? Oh well, i'm probably just being stupid here, i guess.
Basically it is 2x RAID-6 arrays concatenated by RAID0
Yeah, i get the concatenating part but there is no mirroring going on. It's simply making both arrays into one single device. It might be good for speed but data stored doesn't suddenly have more mirrors than before.
If you are lucky enough, you will have 1 2 drives failed at most in each RAID6
UPD. Oops, my bad, forgot about disk quantity in each RAID6
Total drives used for redundancy is doubled. The RAID0 on top of it just joins it together, striping them.
Just like RAID10 is just a bunch of stripes mirrored (mdadm is actually RAID0+1 in a sense) or mirrored pairs striped (HW raid, slower, same redundancy)
Involucration
In general yes but from what i understand a single piece of data is still stored on the same amount of drives as before. It would just ideally end up on one of the 2 underlying arrays. It's not even that important anyways since if a single of the underlying arrays fails the whole thing will be out of service anyways. It's not like data being stored exclusively on a single one of the underlying arrays (which it probably isn't anyways due to - as you already said - striping) would make it somehow (easily) retrievable if the other array fails.
Having 2 (smaller) arrays might make rebuilding faster and less stressful for the disks though, i guess.
Yeah, it's a bit weird. In theory more disks can fail but if your unlucky it still only takes the same amount as before to bring the whole thing down.
This is why monitoring + backups are a thing
Yeah, as i said in the beginning disks usually don't straight up die out of nowhere.
Always have a backup ironicly applies here as it does season to season on LET
No.
You are not looking at the forest from the trees.
Yes, one RAID6 is still limited to 2 max failures before data loss BUT you doubled the total redundancy drives.
6 drives RAID6, it's much less likely 3 drives fail tahn on 12 drives RAID6 before resync completes.
They actually often do.
Well, like i said, kind of but can you really call that redundancy? Maybe the term luck-dependent-redundancy would be more fitting
Out of those few that really died on me most already had suspicious smart readings and if they would have been important to me would have been replaced at that point. I can kind of imagine them dying pretty quickly when stressed with something like a rebuild but otherwise? You've probably seen more drives than i did though.
Isn't RAID1 about (data loss) risk reduction?
Yes. You increased your redundancy factor from 16.67% to 33.33%
Just few tens of thousands.
Most of them fail without warning, with no indication. Most typical indications people use have no bearing on failure rate.
There's been research papers on this, even as of late by backblaze (not sure did they make an actual paper or just released data) showing smart data cannot be used as failure predictor.
Can confirm, I've been through 2-3 RAID5/6 failures with @PulsedMedia
Hmm, i'm not sure. In the simplest case you have two disks that could fail instead of one. Sure it's just half the data that's lost then and in case disk really means another array there is less data that needs to be rebuild, which is faster and less stressful than rebuilding the whole thing but in general i'd say RAID0 is more about speed or convenience of treating two disks as a single device.
Raid 0 is striping. Data is written/read from all the drives at once. It's for speed only and there is no redundancy. If you lose 1 of the drives, you lose the whole array.
OK, maybe that's the accepted math but i still wouldn't feel much safer (outside of faster and less stressful rebuilds).
Interesting. Maybe i was just lucky then.
Yeah, what I meant is RAID1/6 and similar.
not again
It's a bit strange. Where are all the people losing millions right now? There's practically noone but OP and a bunch of nolifes discussing RAID setups in this thread
Well, I think calin is having a bad day right now, tried to reach him, no luck yet.
It has been a bad week for Calin.
I feel bad for @Calin
, so much trouble after his 18th birthday
If you hit rock bottom, the only way is up, so, things might change