Mover very slow

[caplam]

Hi,

it's not a good period for my unraid server.

Having big trouble with write amplification on my cache pool i decided to move from btrfs cache pool to single xfs cache disk until there is a correction.

I shut down docker and vm service.

I'm moving cache share to array with the mover but it only writes at 1,5MB/S. All cache shares (except an empty one ) have been set to cache: yes and included disk:2

Since few days i have so many troubles with this server, that i'm considering moving to another os.

How can i stop the mover and transfer share from cache to array manually ?

I also have a big question. From what i understood docker.img is btrfs filesystem handling cow. When loop2 is mounted we can see /docker/btrfs/subvolumes tree.

 Is this handled correctly on xfs cache ?

godzilla-diagnostics-20200531-1213.zip

[JorgeB]

3 minutes ago, caplam said:

I'm moving cache share to array with the mover but it only writes at 1,5MB/S.

Parity2 appears to be failing, run an extended SMART text to confirm.

 

May 31 10:35:32 godzilla kernel: ata4.00: exception Emask 0x0 SAct 0x1ff80 SErr 0x0 action 0x0
May 31 10:35:32 godzilla kernel: ata4.00: irq_stat 0x40000008
May 31 10:35:32 godzilla kernel: ata4.00: failed command: READ FPDMA QUEUED
May 31 10:35:32 godzilla kernel: ata4.00: cmd 60/00:38:e0:d3:67/04:00:08:00:00/40 tag 7 ncq dma 524288 in
May 31 10:35:32 godzilla kernel:         res 41/40:00:90:d6:67/00:00:08:00:00/00 Emask 0x409 (media error) <F>
May 31 10:35:32 godzilla kernel: ata4.00: status: { DRDY ERR }
May 31 10:35:32 godzilla kernel: ata4.00: error: { UNC }
May 31 10:35:32 godzilla kernel: ata4.00: configured for UDMA/133
May 31 10:35:32 godzilla kernel: sd 5:0:0:0: [sdm] tag#7 UNKNOWN(0x2003) Result: hostbyte=0x00 driverbyte=0x08
May 31 10:35:32 godzilla kernel: sd 5:0:0:0: [sdm] tag#7 Sense Key : 0x3 [current]
May 31 10:35:32 godzilla kernel: sd 5:0:0:0: [sdm] tag#7 ASC=0x11 ASCQ=0x4
May 31 10:35:32 godzilla kernel: sd 5:0:0:0: [sdm] tag#7 CDB: opcode=0x88 88 00 00 00 00 00 08 67 d3 e0 00 00 04 00 00 00
May 31 10:35:32 godzilla kernel: print_req_error: I/O error, dev sdm, sector 141022864
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022800
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022808
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022816
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022824
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022832
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022840
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022848
May 31 10:35:32 godzilla kernel: md: disk29 read error, sector=141022856

 

 

 

 

[caplam]

yes i have read errors on it. So was parity 1 and have it replaced. 

I have no other disk to replace parity 2. Can i unassigned it  and go on with only 1 parity drive ?

 

edit: i'm running a smart test. Last week one was ok.

Edited May 31, 2020 by caplam

[JorgeB]

3 minutes ago, caplam said:

Can i unassigned it  and go on with only 1 parity drive ?

Yep.

[caplam]

ok. I can't because mover is running. 

Can we stop it ?

 

edit : simple as : mover stop

Edited May 31, 2020 by caplam

[caplam]

i unassigned parity2, restarted the array and invoked mover and.....

speed is the same between 2 and 5 MB/S. 

There are no other writes except disk2 and parity1. It should be quicker . 

[JorgeB]

21 hours ago, caplam said:

speed is the same between 2 and 5 MB/S. 

You're likely moving a lot of small files, mover is not efficient for that, better off moving manually.

[caplam]

that's what did.

Mover took 3 hours to move appdata/plex (13G).