[HOWTO] How to increase raid rebuild speed

[scotland21]

I've finally figured it out, I was using the command to see the max and min instead of the echo command to set the speed.
Thanks for the reply don.

[scotland21]

This doesn't appear to work on the ts-410, I can confirm the new speed settings but the transfer rate doesn't improve.
Any ideas ?

[Don]

What is the CPU utilization?

[scotland21]

Between 33 and 46 percent and around 175 mb free ram, this doesn't change from the default min of 1000 and any higher speeds I set.
I do have around 1.7 tb of data on the 3 drives and they are hd films, don't know if this makes a difference.
The write speed is anywhere between 4000K /sec and 6000K /sec and at 64% I still have approx 1600 mins left and this is after almost 48 hrs now.

[Don]

What do the following show?

cat /proc/sys/dev/raid/speed_limit_max
cat /proc/sys/dev/raid/speed_limit_min
cat /proc/mdstat

[scotland21]

cat /proc/sys/dev/raid/speed_limit_max is 200000
cat /proc/sys/dev/raid/speed_limit_min is 11000
cat /proc/mdstat write rate 4812K /sec at 65% with 1767 min left

[Don]

Increase the min speed.

[scotland21]

that's the problem, no matter what speed I set 1000 to 120000 the cpu utilization and transfer rates do not change
I wonder if there is a limit on ts-410

[Don]

Well either QNAP is limiting it in some other way or you have hit the max speed of the hardware.

[lifeisfun]

I experience exactly the same problem on my 509. I added 1.5TB HD to my 3x 1.5tb ( + 1 x 1.5TB spare) Raid 5
10 hours ago and progress of expanding is only 6%
CPU usage never goes over 5%
When I increased rebuild speed to 50000 as recommended by OP I get CPU usage about 7%
Is there any risk to try let's say 190000 ?

Thanks

System InformationCPU Usage 3.6 %

Total Memory 3541.8 MB


Free Memory 3424.5 MB


Ethernet1 Packets Received 220799

Ethernet1 Packets Sent 140468

Ethernet1 Error Packets 0

Ethernet2 Packets Received 1548551

Ethernet2 Packets Sent 6283927

Ethernet2 Error Packets 0

CPU Temperature 28°C/82°F

System temperature 36°C/96°F

HDD 1 temperature 36°C/96°F

HDD 2 temperature 36°C/96°F

HDD 3 temperature 36°C/96°F

HDD 4 temperature 36°C/96°F

HDD 5 temperature 36°C/96°F

HDD 6 temperature

HDD 7 temperature

HDD 8 temperature

System Up Time 1 Day 14 Hour 53 Minute(s)


System fan speed 869

[lifeisfun]

Setting speed to 190000 bumped the CPU to +/- 7.6%


CPU Usage 7.6 %

Total Memory 3541.8 MB


Free Memory 3424.4 MB


Ethernet1 Packets Received 243397

Ethernet1 Packets Sent 154721

Ethernet1 Error Packets 0

Ethernet2 Packets Received 1572515

Ethernet2 Packets Sent 6307341

Ethernet2 Error Packets 0

CPU Temperature 28°C/82°F

System temperature 37°C/98°F

HDD 1 temperature 37°C/98°F

HDD 2 temperature 36°C/96°F

HDD 3 temperature 37°C/98°F

HDD 4 temperature 37°C/98°F

HDD 5 temperature 36°C/96°F



System Up Time 1 Day 18 Hour 50 Minute(s)


System fan speed 869

[bpwg549]

On a Ts-859u, (6) 2TB drives, Raid 6 and one hot spare I've got a new rebuild (scratch) down to about 12-14 hours. I've got the minimum (above) set to 25,000 and the max set to 50,000. Cpu is running almost 100%. I put the fan on manual (fast), which is keeping it cool. On auto, it was getting pretty hot, before the fans would ramp up.

The above settings seem to be pretty variable as to the speed of the cpu, etc.

[ibob]

I have followed the tips in this thread, but my NAS is still extremely slow.
I have a TS-409 with firmware 3.1.0 Build 0708T

Until now I have run my NAS as JBOD, but since a recent disk crash I decided to buy new HDDs and set up a Raid, which I haven't tried before.

I've installed 4 brand new Seagate Barracuda 7200.12 1TB HDDs (ST3100528AS with firmware CC44).

I've collected all my data (1 TB) to Drive 1 and made Drive 2+3+4 into a Raid 5 (which took, cant remember exactly, but more than 24 hours).

Then moved all data from Drive 1 to the Raid (which took exactly 25 hours and 27 minutes).

Just for fun I tried to transfer a 4 GB file from my iMac (Samba) to the NAS via a direct ethernet cable, which usually takes around 8 minutes to a Single Disk Volume. It said transfer time approximately an hour?! - so I cancelled it. I rebooted the NAS but that didn't improve the transfer speed, so I dropped it.

Now Im adding Drive 1 to the Raid, so I will end up having a Raid 5 with 4 drives in total.

I started at 2:40 PM today and now at 11:11 PM it's at 10%.

Running cat /proc/mdstat
- the speed says 6224K/sec and finish=2778.1min

I understand that this process might take a while, but an estimated finish in 2778.1 minutes?!
That's more than 46 hours?!

The CPU is around 60%.

I've upped the speed_limit_min from 1000 to 50000 but it doesn't seem to change the speed.

cat /proc/sys/dev/raid/speed_limit_max
200000
cat /proc/sys/dev/raid/speed_limit_min
50000

Is this the speeds I can expect from a Raid 5 with four 1TB disks on my NAS? Because then Im seriously considering going back to JBOD, even if I have the risk of another crash.
I can live with the building of the Raid takes a long time, since Im (hopefully) only doing it once. But write speeds at 1MB/sec when transferring files from my computer to the NAS is not acceptable.

Thanks for any advice.

[Don]

Transfering while the raid is building/rebuilding will be very slow.

[Tarrant1701]

Just wanted to share my experiences. Have a TS-509 Pro that had 750GB x5 drives that I expanded with 2TB WD20EARS drives. After replacing each drive sequentially, I am now waiting for the RAID to complete expansion. I found this thread because the expansion process was S-L-O-W!

Using Don's excellent tip, I found the following:
set_limit_min was default at 5000
set_limit_max was default at 200000

Under default conditions, I was getting "spikes" in CPU usage rather than a constant level of CPU activity. Peaks hit perhaps 30% followed by valleys of near-0% usage. The CPU usage monitor looked like a saw-tooth pattern because of this.

I used the ECHO command posted by Don to set_limit_min to 50000. After that, I got a constant level of CPU activity of about 50%. I tried increasing the MIN even higher, to 75000, 100000, or even 150000 without any further increase in CPU usage. Makes me think I hit the bottleneck imposed by the drives themselves, since these are "Green" drives that I am using. Still, I'm happy for the increase in rebuild speed, yay! I would up setting for a MIN of 750000.

One question though: in the web interface, the RAID Management screen shows that my current rebuild is at 32%. However, the cat /proc/mdstat command shows:

resync = 64.7$ finish = 224.8 min speed 50974K/sec

Any idea why there is a big discrepancy between what mdstat reports and what the RAID Management web interface shows?