Storage setup recommendation

[luca1001]

Hi all,

I'm setting up freenas 11.3 on a Dell R620, server has the following specs:
- 2x E5-2670 / 128 GB RAM
- dual SFP+ / dual I350 daughter card
- H710 controller in IT mode
- 8x MX500 1TB SSD drives

The main purpose of this server is to provide iSCSI storage to a separate ESXi 6.7 host via the dual SFP+ ports. ESXi hosts a couple of windows VMs (low load), along with 5/6 linux VMs ubuntu 18.04.3 LTS that can have quite intensive I/O (database) load. The linux VMs are similar as far as OS type / configuration. I will backup the VMs periodically, and downtime for rebuilding after a failed SSD drive is not a big deal.

Being new to freenas, I'm not sure how to setup storage.

- I've been reading this thread but having a hard time translating it to concrete storage configuration): should should I use some kind of mirror-based setup, or a single raid-z / zvol at 50% capacity?
- should I enable de-duplication?
- SLOG, ZIL, L2ARC?

 

[danb35]

should I use some kind of mirror-based setup

Yes. Four vdevs of two mirrored disks each.

should I enable de-duplication?

No.

SLOG, ZIL, L2ARC?

iSCSI doesn't use sync writes by default, so a SLOG would actually slow down the system unless you changed that default. I'd wait on L2ARC until you see you need it--the system monitoring will tell you the ARC hit ratio.

 

[luca1001]

Yes. Four vdevs of two mirrored disks each.

No.

iSCSI doesn't use sync writes by default, so a SLOG would actually slow down the system unless you changed that default. I'd wait on L2ARC until you see you need it--the system monitoring will tell you the ARC hit ratio.

Thanks danb35, that's what I'll do. Do you think it's a valid test to setup a VM on the iSCSI datastore and run a disk benchmark like crystalmark? I've run such tests with both 8 drives in raid-z2 and the 4 vdevs mirror.

 

[luca1001]

After a few weeks of running freenas I had a lok at ARC hit ratio.

Do I read this correctly, that L2ARC is not needed in my case?

 

[JaimieV]

Yes, no L2ARC needed. Your hit ratio is splendidly high. Congratulations!

 

[HoneyBadger]

L2ARC is even less necessary here because your pool is all-flash; to be beneficial, L2ARC has to be faster than your main pool.

Regarding SLOG/sync writes, as long as you are willing to restore from backup you don't technically need it; but how crucial is the data in those databases?

 

[luca1001]

L2ARC is even less necessary here because your pool is all-flash; to be beneficial, L2ARC has to be faster than your main pool.

Regarding SLOG/sync writes, as long as you are willing to restore from backup you don't technically need it; but how crucial is the data in those databases?

There is a UPS that should take the server down gently in case of power troubles, and yes, restoring from backup should be fine.

On a separate note, I'm getting plagued by some errors, since updating to 11.3-U1 (I had no errors with 11.3):

CRITICAL
Device: /dev/da4 [SAT], 1 Currently unreadable (pending) sectors

Every time the error is about different drive (averaging about one error / day). I doubt this is an actual problem with my hardware / cabling. I've seen this could be caused by some sort of firmware compatibility issue between Freenas and the MX500 SSD units, which is what I'm using.

 

[JaimieV]

Please check your SMART details. In a shell, smartctl -a /dev/da4 (or whichever others), and look in the "SMART Attributes Data Structure" block for these in particular:
5 Reallocated_Sector_Ct 0x0033 100 100 050 Pre-fail Always - 0
196 Reallocated_Event_Count 0x0032 100 100 000 Old_age Always - 0
198 Offline_Uncorrectable 0x0030 100 100 000 Old_age Offline - 0
Zero at the end is the good result.

 

[HoneyBadger]

There is a UPS that should take the server down gently in case of power troubles, and yes, restoring from backup should be fine.

UPS solves a different problem than SLOG/sync writes, but if restore from backup is acceptable then that ultimately answers it - just make sure that any automated shutdown sequence due to power loss brings the hypervisors down first.

Every time the error is about different drive (averaging about one error / day). I doubt this is an actual problem with my hardware / cabling. I've seen this could be caused by some sort of firmware compatibility issue between Freenas and the MX500 SSD units, which is what I'm using.

This is a firmware bug with the MX500 (and possibly other Crucial drives) that apparently isn't ever going to be fixed, since they consider it as "operating normally."

Current Pending Sector Count warnings on SSD

Hello, I've been receiving a ton of warnings about Current Pending Sector Count ticking up to 1, and then back down to zero on my Samsung SSD. The SSD was quite old and too small for my current needs, so I just replaced it with a new Crucial SSD. The problem persisted with the new drive, so I rep...

forums.unraid.net