RAID rebuild time

[tikker]

Is there a way to estimate how long it will take to rebuild a RAID 5 or 6 array? Since the change of failure is highest at these times, I was wondering, when I experience a drive failure in a 5x4TB RAID 5, for example, how long a rebuild would take.

Is it just rebuild time = (HDD size / HDD speed) with a bit of overhead for parity in the most ideal case?

[KirbyTech]

7 minutes ago, tikker said:

Is there a way to estimate how long it will take to rebuild a RAID 5 or 6 array? Since the change of failure is highest at these times, I was wondering, when I experience a drive failure in a 5x4TB RAID 5, for example, how long a rebuild would take.

Is it just rebuild time = (HDD size / HDD speed) with a bit of overhead for parity in the most ideal case?

There is no real way to know for sure, partly depends on your OS and tons of other things. When I calculate it, I generally go with SIZE/(SPEED*0.75)=TIME

[Razor Blade]

There are a lot of variables... A few variables would be type of RAID, RAID or HDD controller, file system used, how much data on disks, drive type and speed (not just platter speed), if you're reading/writing data while the array is rebuilding, etc.

 

[Windows7ge]

I had to resilver my server 8 drive raid6 and I factored if I had 1TB of data/8 = ~ 128GB / 175MB/s = seconds to resilver array...rough estimate.

[Electronics Wizardy]

5 minutes ago, Windows7ge said:

I had to resilver my server 8 drive raid6 and I factored if I had 1TB of data/8 = ~ 128GB / 175MB/s = seconds to resilver array...rough estimate.

Im guessing your using ZFS

 

Though one thing to note with zfs is a rebuilt or resilver only moves where there is data on the drive, unlike with hardware raid or other software raid solutions like md where they will always rebuild the whole drive, no matter how much data is on it.

[Windows7ge]

2 hours ago, Electronics Wizardy said:

Im guessing your using ZFS

Weather you're basing that presumption off of past conversations we've had or from the keyword "resilver" you are correct. To be more specific it's ZFS on FreeNAS.

 

To expand on your explanation I do have to ask does that mean when the Partition/RAID is first constructed does it format the entire disk(s) or rather does it expand the partition as data is added?

[Electronics Wizardy]

Just now, Windows7ge said:

To expand on your explanation I do have to ask does that mean when the Partition/RAID is first constructed does it format the entire disk(s) or rather does it expand the partition as data is added?

It formats the whole new drive, but only copyies the blocks with data over, as ZFS knows what block have data, unlike hardware raid that doesn't know where the data is.

[tikker]

On 07/04/2018 at 5:53 PM, KirbyTech said:

There is no real way to know for sure, partly depends on your OS and tons of other things. When I calculate it, I generally go with SIZE/(SPEED*0.75)=TIME

Ah kind of what I was expecting.

On 07/04/2018 at 5:54 PM, Razor Blade said:

There are a lot of variables... A few variables would be type of RAID, RAID or HDD controller, file system used, how much data on disks, drive type and speed (not just platter speed), if you're reading/writing data while the array is rebuilding, etc.

 

 

On 07/04/2018 at 5:55 PM, Windows7ge said:

I had to resilver my server 8 drive raid6 and I factored if I had 1TB of data/8 = ~ 128GB / 175MB/s = seconds to resilver array...rough estimate.

It won't be used when rebuilding most likely (it's not used that intensively in the first place).

 

So rough estimates is all we'll get. That works for me. Hopefully it'll be a long time before this happens and fingers crossed stuff doesn't die while rebuilding  

[jde3]

This is actually a pretty big topic, but.. software to the rescue, I believe newer versions of ZFS will just tell you. do zpool status. it should tell you when it started and give you some indication of its progress and speed. (how accurate it is idk)

 

resilver on zfs is basically a scrub. so as long as it takes to scrub should be the same as resilver I would think.

[tikker]

10 hours ago, jde3 said:

This is actually a pretty big topic, but.. software to the rescue, I believe newer versions of ZFS will just tell you. do zpool status. it should tell you when it started and give you some indication of its progress and speed. (how accurate it is idk)

 

resilver on zfs is basically a scrub. so as long as it takes to scrub should be the same as resilver I would think.

In the end this probably ties back to the age old problem of "predicting the unpredictable" of progress bars, I guess. I was thinking about using Unraid though, so no ZFS for me.

[jde3]

3 minutes ago, tikker said:

In the end this probably ties back to the age old problem of "predicting the unpredictable" of progress bars, I guess. I was thinking about using Unraid though, so no ZFS for me.

Ahh your killin me smalls.

 

Pretty sure Linus is the only person that's ever used Unraid for anything.. but you know.. it's your data.

[tikker]

12 minutes ago, jde3 said:

Ahh your killin me smalls.

 

Pretty sure Linus is the only person that's ever used Unraid for anything.. but you know.. it's your data.

It's either unRAID or RAID 5. The formers expandability sounds nice, but on the other hand just setting up a proper RAID 5 once sounds much better and interesting, so who knows First I'll have to finish getting all the drives.

[jde3]

Expandability isn't everything. It sounds nice, "hea I'll just pop in more drives" but in practice there is a limit to how many disks you can actually attach to a system and what you might just want in the future is just larger newer drives, something ZFS does just fine.

 

I don't want to start a flame war, you can use unraid/btrfs and it's fine for home use sure, anything works for home but it isn't really accepted in industry (to my knowledge at least) and I'm not sure it ever will be due to it's rough start and potential design problems, that could change and I could be totally wrong.. it's just.. usually not the route to success and acceptance file systems take though (it reminds me of ReiserFS 10 years ago). As a sysadmin I haven't spent any time learning it because.. I honestly never see it. I work and have worked generally with Sun, FreeBSD, Linux, some OpenBSD and Unix systems in general though so maybe I'm a self selecting sample here due to my specialization in the industry..

 

From a technical aspect btrfs is a cool idea.. they have to get the design right though. That is NOT easy many large corporations and brilliant people have failed down this road.

[NelizMastr]

Generally speaking, the likelyhood of a 2nd drive crash during a rebuild is rather slim, at least if you didn't buy consumer drives from the same batch. NAS and Enterprise drives are always picked from different batches when you order them, which minimizes the risk of a faulty batch causing massive data loss.

[tikker]

28 minutes ago, NelizMastr said:

Generally speaking, the likelyhood of a 2nd drive crash during a rebuild is rather slim, at least if you didn't buy consumer drives from the same batch. NAS and Enterprise drives are always picked from different batches when you order them, which minimizes the risk of a faulty batch causing massive data loss.

I was wondering about the batches if I were to order 6 drives at once. That's cool if they already take that into account with NAS drives.

[NelizMastr]

3 minutes ago, tikker said:

I was wondering about the batches if I were to order 6 drives at once. That's cool if they already take that into account with NAS drives.

Yep, a decent retailer will do this for you (as it's per instruction of the harddisk manufacturers). My 2 4TB Ironwolf drives are from completely different batches and were ordered at once.

[jde3]

It can happen, don't think that it *never* will. Backups are still necessary. (ie: having the data in two places)

[NelizMastr]

1 hour ago, jde3 said:

It can happen, don't think that it *never* will. Backups are still necessary. (ie: having the data in two places)

Oh no doubt. RAID is never a back-up.

[Jarsky]

11 hours ago, tikker said:

It's either unRAID or RAID 5. The formers expandability sounds nice, but on the other hand just setting up a proper RAID 5 once sounds much better and interesting, so who knows First I'll have to finish getting all the drives.

 

Are you saying hardware raid isn't expandable? 

[tikker]

3 minutes ago, Jarsky said:

 

Are you saying hardware raid isn't expandable? 

Not easliy right?

 

Anyway, after seeing @AshleyAshes topic, FlexRAID has now caught my attention  

[Jarsky]

Hardware RAID is typically quite easy. Especially LSI with their MegaRAID software, its probably the easiest. 

e.g http://sprava-site-stanic-a-serveru.cz/lsi-megaraid-manager-raid-array-extension.htm

 

Of course Hardware RAID is more expensive