SOLVED What's causing a drop of performance ???

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
I rarely transfer a lot of data but tonigh, I ripped some DVDs and BD and when I wanted to transfer them to my NAS, I realized that the transfer speed for a single file started at 100 MB/s, which seems normal for a 1Gb/s NIC, but dropped at 15-25 MB/s which I find really slow. The scenario repeats itself for each single file.

Also, if I try to transfer a batch of 2-3 files, then the write speed is below 15MB/s for each file.

How can I investigate to source of that: network, lack of memory for buffer ???
 

ChrisRJ

Wizard
Joined
Oct 23, 2020
Messages
1,909
How are your disks connected?

To separate the network from the storage dimension you should test them separately. For the network side I would use "iperf" and for storage "fio".
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
How are your disks connected?

To separate the network from the storage dimension you should test them separately. For the network side I would use "iperf" and for storage "fio".
I am not sure or to perform th "fio" test. I would not want to mess all my system performing a test. I have found this:

But a more directed help would be appreciated if possible.
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
Anyone would follow on this please?
 

ChrisRJ

Wizard
Joined
Oct 23, 2020
Messages
1,909
I am not sure or to perform th "fio" test. I would not want to mess all my system performing a test.
Can you elaborate what the problem is?
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
Can you elaborate what the problem is?
Hello Chris,
As stated in my original post, if I transfer a single file, the datarate from my PC to my TN Core is around 115MBs which I believe is normal on a Gb lan.

I just ripped three 4K movies and started the transfer to my media library.
After about 3 minutes of 115 MBs of sustained througput, the transfer has dropped to around 15MBs and I don't know how to isolate the problem.
 

Samuel Tai

Never underestimate your own stupidity
Moderator
Joined
Apr 24, 2020
Messages
5,399
How is your pool configured, and how much RAM do you have in your system?
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
How is your pool configured, and how much RAM do you have in your system?
I think most of the info is in my signature.
What else can I provide concerning the pool? Is there an "ouptput" I could use to best describe the thing ?
I just ran a backup from a Proxmox VM on an NFS share and it was even slower.
 

Samuel Tai

Never underestimate your own stupidity
Moderator
Joined
Apr 24, 2020
Messages
5,399
What's the output of zpool status <name of your pool>? How full is your pool? zpool get capacity <name of your pool>.
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
What's the output of zpool status <name of your pool>? How full is your pool? zpool get capacity <name of your pool>.
Code:
truenas% zpool status Bassin
  pool: Bassin
 state: ONLINE
  scan: scrub repaired 0B in 1 days 02:32:52 with 0 errors on Thu Feb  2 04:33:34 2023
config:

        NAME                                            STATE     READ WRITE CKSUM
        Bassin                                          ONLINE       0     0     0
          raidz2-0                                      ONLINE       0     0     0
            gptid/3fc56edc-7243-11eb-a7fc-842b2b51538a  ONLINE       0     0     0
            gptid/b84df4dd-f29a-11ec-98d9-842b2b51538a  ONLINE       0     0     0
            gptid/3ffedd3a-7243-11eb-a7fc-842b2b51538a  ONLINE       0     0     0
            gptid/41060316-7243-11eb-a7fc-842b2b51538a  ONLINE       0     0     0
            gptid/4131db70-7243-11eb-a7fc-842b2b51538a  ONLINE       0     0     0

errors: No known data errors


Code:
truenas% zpool get capacity Bassin
NAME    PROPERTY  VALUE  SOURCE
Bassin  capacity  31%    -
 

John Doe

Guru
Joined
Aug 16, 2011
Messages
635
maybe check with iperf if your network is having issues
check cpu load while transfer on client and server
assuming you use windows, maybe check task manager -> performance CPU, network, Disk business
what nic is in your client?

looking at your signature, i am a bit confused, you are mixing 2 and 4 tb drives?
 

Samuel Tai

Never underestimate your own stupidity
Moderator
Joined
Apr 24, 2020
Messages
5,399
OK, what you're seeing is the write bandwidth of the slowest drive in your RAIDZ2 pool. You may benefit by adding a log SSD to your pool, to offload the write log from your pool disks. Note, your log SSD should have power loss protection. Optanes make good log VDEVs.
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
looking at your signature, i am a bit confused, you are mixing 2 and 4 tb drives?
I made my initial install with 4X 2TB HDD that were already in my Windows PC plus 1 new 4TB knowing that I would not get the full capacity until all 5 disks were 4TB. One 2TB has since showed SMART warning and have been upgraded to 4TB.

OK, what you're seeing is the write bandwidth of the slowest drive in your RAIDZ2 pool. You may benefit by adding a log SSD to your pool, to offload the write log from your pool disks. Note, your log SSD should have power loss protection. Optanes make good log VDEVs.
How can it sustain the througput for a few minutes and not just a few seconds ?

What would be a suggested capacity for such an Optane? Please consider all 5 drives are 4TB which is where it will get to.
 

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
OK, what you're seeing is the write bandwidth of the slowest drive in your RAIDZ2 pool. You may benefit by adding a log SSD to your pool, to offload the write log from your pool disks. Note, your log SSD should have power loss protection. Optanes make good log VDEVs.
I sadly realized that although I only ordered Red Pro drives, one of them is an SMR drive: WD40EFAX

Can it be source of the problem?
 

Samuel Tai

Never underestimate your own stupidity
Moderator
Joined
Apr 24, 2020
Messages
5,399

NinthWave

Contributor
Joined
Jan 9, 2021
Messages
129
Yes, that's the drive with the lowest write bandwidth.
This drive is not that old in terms of hours but if I remember my readings about SMR, they are to be avoided at all costs...
So I guess I should be replacing this drive instead of going the Optane's log way ?
 
Joined
Oct 22, 2019
Messages
3,589
So I guess I should be replacing this drive instead of going the Optane's log way ?
To safely replace that drive (with a CMR) is definitely your best bet.

FYI: Your signature is not correct. You don't have all Red "Pro" drives. The WD40EFAX is a 4TB WD Red. (Not Red Plus. Not Red Pro.)
 
Top