RAID recovery time is in the days?

[BadBoyHaloCat]

This is my first RAID 5 array (3x2TB) and I'm initializing it with software RAID (mdadm) and the ETA for the initial rebuild (to initialize parity) is about 1.5 days. Normal?

[Electronics Wizardy]

What drives are you using? What speeds do you see for the initialization? 

 

Since your on linux, id personally go with zfs for most uses as its better at protecing data, and support things like snapshots much better.

[BadBoyHaloCat]

2 minutes ago, Electronics Wizardy said:

What drives are you using? What speeds do you see for the initialization? 

 

Since your on linux, id personally go with zfs for most uses as its better at protecing data, and support things like snapshots much better.

I was going to do btrfs because I prefer it, what are the benefits of zfs over btrfs?

 

I'm getting about 12MB/s. Here's mdstat with the speeds etc:

Personalities : [raid6] [raid5] [raid4] 
md0 : active raid5 sdc1[3] sdb1[1] sda1[0]
      3906557952 blocks super 1.2 level 5, 512k chunk, algorithm 2 [3/2] [UU_]
      [>....................]  recovery =  1.5% (30093780/1953278976) finish=2526.7min speed=12685K/sec
      bitmap: 0/15 pages [0KB], 65536KB chunk

unused devices: <none>

I'm using 3 of the Seagate BarraCuda 2TB Internal Hard Drive (3.5 Inch SATA 6Gb/s 7200 RPM 256MB Cache) drives.

[Electronics Wizardy]

Just now, BadBoyHaloCat said:

was going to do btrfs because I prefer it, what are the benefits of zfs over btrfs?

Generally more stable, and since zfs is doing raid it can use checksums for fixing file integrity, doesn't require an initialization, only rebuilds data. Nice cmd interface. Thats what Id use here.

 

2 minutes ago, BadBoyHaloCat said:

I'm using 3 of the Seagate BarraCuda 2TB Internal Hard Drive (3.5 Inch SATA 6Gb/s 7200 RPM 256MB Cache) drives.

Thats your problem. Those are smr drives and will be very slow in raid.

 

Id reaally just get a single larger hdd here instead of raid. Why not just get a single 4tb hdd.

[BadBoyHaloCat]

That makes more sense. I'll consider zfs and maybe a larger single drive.

[BadBoyHaloCat]

Welp. It's 32-bit so I think zfs won't be a good choice anyway.

[Nick7]

Did you verify /proc/sys/dev/raid/speed_limit_max ?

It limits rebuild speed.

Try setting it higher:

# echo 200000 > /proc/sys/dev/raid/speed_limit_max

 

You can also try following:

# echo 4096 > /sys/block/md0/md/stripe_cache_size

[NelizMastr]

SMR will never be fast. You'll get 30MB/s in a RAID if you're lucky, but not much more, even with 8TB drives.

[BadBoyHaloCat]

I'll just use these drives for backup (as that would be sequential writes instead of random access).

During a mdadm resync (if I used RAID 1), would the SMR drives be written to in a consecutive pattern or would it write randomly (if that makes any sense)?

[Electronics Wizardy]

6 hours ago, BadBoyHaloCat said:

I'll just use these drives for backup (as that would be sequential writes instead of random access).

During a mdadm resync (if I used RAID 1), would the SMR drives be written to in a consecutive pattern or would it write randomly (if that makes any sense)?

It should be sequntical for rebuilds in raid 1. But SMR drives often aren't great at writes, but this should be fine. 

[BadBoyHaloCat]

On 6/7/2021 at 9:08 PM, Electronics Wizardy said:

It should be sequntical for rebuilds in raid 1. But SMR drives often aren't great at writes, but this should be fine. 

The initial resync was about 1.2 days, but after that it works fine in RAID 1 for my usage. Thank you for the help.