Resilver speed - slow?

billgreenwood

Dabbler
Joined
Apr 11, 2014
Messages
42
Hi all,

I'm resilvering a disk it's and it's extremley slow. I dont have any users connected, i have AFP and SMB, Smart and UPS on. Nothing else, no jails running, no VMs, no plugins.

I'm getting 27.8M/s

I have
vfs.zfs.resilver_delay=0
vfs.zfs.resilver_min_time_ms=500
vfs.zfs.scrub_delay=0
vfs.zfs.top_maxinflight=128

My system is in my sig, I know it's not the fastest but the crux of it is
System - FreeNAS-11.3-RELEASE
Motherboard: Supermicro MBD-X9SCM-F-O
CPU: Intel BX80637E31240V2 Xeon Quad-Core Processor 3.4 GHz LGA1155
RAM: 32GB DDR3 PC3-10600
SATA Controller: IBM M1015 RAID card flashed to IT mode
Disks: 10 x WD Red 6TB WD60EFAX (Raid z2)


This is from my zpool status | more command
pool: zpool
state: DEGRADED
status: One or more devices is currently being resilvered. The pool will
continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
scan: resilver in progress since Tue Mar 10 23:41:51 2020
2.12T scanned at 1.02G/s, 57.4G issued at 27.8M/s, 42.2T total
3.97G resilvered, 0.13% done, 18 days 09:50:53 to go
config:

NAME STATE READ WRITE CKSUM
zpool DEGRADED 0 0 0
raidz2-0 DEGRADED 0 0 0
gptid/d440d1bb-5ba1-11e6-8771-002590db3d35 ONLINE 0 0 0
gptid/6b71bf82-caa0-11e5-a2cf-002590db3d35 ONLINE 0 0 0
gptid/6c85a56d-caa0-11e5-a2cf-002590db3d35 ONLINE 0 0 0
gptid/9866ea39-b6be-11e9-83ae-001b21be2188 ONLINE 0 0 0
gptid/6ea11b8e-caa0-11e5-a2cf-002590db3d35 ONLINE 0 0 0
gptid/6faacfb8-caa0-11e5-a2cf-002590db3d35 ONLINE 0 0 0
gptid/a86075cf-50b2-11e6-88c5-002590db3d35 ONLINE 0 0 0
replacing-7 DEGRADED 0 0 0
2938101774742790080 UNAVAIL 0 0 0 was /dev/gptid/e216a300-dfa5-11e5-a1df-002590db3d35
gptid/ac979e39-5f0f-11ea-b360-002590db3d35 ONLINE 0 0 0
gptid/721b320a-caa0-11e5-a2cf-002590db3d35 ONLINE 0 0 0
gptid/732f8318-caa0-11e5-a2cf-002590db3d35 ONLINE 0 0 0

errors: No known data errors


Can anyone see any issues here?

Or have any input or ideas



Cheers

Bill






(I resilvered a disk on another system last week and the same WD RED 6tb drive took about 12 hours)
My other system is
Supermicro MBD-X10SL7-F-O
Intel(R) Xeon(R) CPU E3-1220 v3 @ 3.10GHz (4 cores)
RAM: 32GB DDR3 PC3-10600
LSI 2308 flash to IT mode
Disks: 10 x WD Red 6TB WD60EFAX (Raid z2)
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,681
The first portion of a resilver is usually very slow due to a lot of it being metadata, and then the rest can be slow if you have a lot of small files. I wouldn't be concerned by claimed resilver times during the first few percent of a resilver.
 

hervon

Patron
Joined
Apr 23, 2012
Messages
353
Searched the forums ?
 

billgreenwood

Dabbler
Joined
Apr 11, 2014
Messages
42
I didn't check for drive type! It's something I didn't know/consider. This is why this community is so good!!!!

All my other drives are WD60EFRX. Just this single new drive is WD60EFAX !
 
Last edited:
Joined
Jan 27, 2020
Messages
577
I'm kinda baffled that there exists shingled NAS-drives...
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,681
I didn't check for drive type!

Don't feel bad, it's been tripping me up too that I keep forgetting to check model numbers.

I'm kinda baffled that there exists shingled NAS-drives...

... because, why, exactly? :smile:

See this is the thing. With the advent of reasonably affordable SSD, applications that need data quickly are being serviced that way, and HDD is headed towards simple bulk storage, which is basically always where it's been headed. "NAS" has become more of a synonym for "home NAS" and that's a market where I expect companies expect the yokels, that is, their nonprofessional customers, to be tolerant of lesser performance as long as the NAS actually stores stuff eventually.

Of course for ZFS the problem is that shingled ends up being catastrophically slow, which probably isn't the case for many NAS applications where you just have a busybox Linux storing data sequentially on ext3, so the SMR NAS HDD is merely slower there on a conventional NAS HDD because it is overwriting whole stripes of data at a time, not doing the small-block updates ZFS will tend to do on a fuller pool.

I would be expecting to see all HDD's trending strongly towards any technology that writes slower but increases density, and I hope I've kinda explained why I think that. This sucks for ZFS.
 

billgreenwood

Dabbler
Joined
Apr 11, 2014
Messages
42
These WD60EFRX drives are difficult to get quickly. The Western Digital Reds 6tb WD60EFRX drives are quite difficult to get hold off, in fact they are listed as EOL - https://www2.iosafe.com/support-nas-compatibility-eol-drives - I've just ordered a replacement from Italy. Thinking aloud I may have to make a new upgraded freenas. @jgreco Can you give me a suggestion of what disks to go for next time around? I will probably start replacing my 6gb drives with 14gb. - this looks a decent article. https://www.anandtech.com/show/12075/best-consumer-hdds Hopefully I will not need to change another drive for a while
 
Top