HDD Firmware Update and Sectors

TrueNAS General
1

Hello Guys,

So, now that i’ve got all the drives (16*Seagate EXOS X16 16TB SATA), i’ve got a few questions. This is going to be in a new NAS and so will be the pool. I plan to use RAID-Z2 setup (2 VDEVs, 8 disks each). I’ve a few questions:

Is this pool layout fine? I choose this layout after reading several threads and this seems to have a balance between redundancy and capacity.

My X16 drives are having the SN03 firmware and on the Seagate official site, i see that there are updates available for my drives and the latest firmware is SN04. Although, i cannot find any release notes as what it aims to fix or simply the changelog, do you think updating the firmware will benefit? I’m also bit afraid by turning the drive inoperable as mentioned by a few users. Moreover, should i really update the firmware on all the drives or use as is supplied (SN03)?

Lastly, as per my check, these drives seems to be 4Kn but is formatted in 512e (factory default). So should i convert/format the drive to 4Kn before the deployment? Will changing the sector to 4K yield any better results as compared to 512e?

If so, what should i do first?

Update the firmware, change the sector, do burn in test
Change the sector, update the firmware, do burn in test

Please let me know. Any suggestion is highly appreciated!

Thanks

2

A bit wasteful to do a burn-in test before changing the sector size and then again after the change.

EDIT: I misread this part. I don’t think the order matters too much, but I updated the firmware before changing the sector size, for what it’s worth.

You probably won’t see any real benefits, unless they fixed bugs.

I went ahead and updated the firmware on my Exos and formatted them to 4Kn just because I could. I had no reason to, but I figured I might as well do it now since it doesn’t take long.

I guess there’s a risk if you suddenly lose power during the update or format.

What you really need to change for the Exos is the aggressive power saving features. Not sure if it applies to the X16’s, but my X20’s park the heads way too often. I permanently disabled all power saving features.

People will argue that these enterprise drives are designed to handle a billion load cycles, but I don’t care. The savings on electricity are marginal, and the only real rationale I could find from Seagate had little to do with drive health and more to do with “energy efficiency”.

3

So, i guess update the firmware first, change the sector size and then do the burn in test, yeah?

Yeah, but either way, its always better to have the latest firmware to avoid possible bugs i guess.

I heard that the 4Kn will perform better and probably have better write speeds as compared to 512e. Is that true?

I also read somewhere about ashift=9 and ashift=12. The 9 for the 512e and the 12 for the 4Kn. So, will TrueNAS take care of it automatically when creating a pool or will i have to configure these parameters?

Oh yeah, right right.

OMG. I was not aware of this. I thought TrueNAS always set the disk as active and it does not go idle. But i think my X16 do go to kind of power saving mode as sometimes i have noticed that if i try to access the pool after a while, it will take a moment and then i can access the pool contents. I think this is why the WD has some power saving disabled enterprise drives?

Also, by the power saving, do you mean drive motor stops spinning or park the head after a certain period of time defined/set by the manufacturer?

Can you tell me how to disable the power saving features on the Seagate drives? Is that some kind of setting that can be done within the TrueNAS or the BIOS?

Hmm. So, running the drive always instead of letting it go in idle power saving mode is beneficial in terms of drive health and durability. Is that what you mean?

4

Probably not enough to notice. Supposedly on single threads, there’s a benefit, but not for multithreaded operations.

TrueNAS uses ashift=12 by default. It has for a long time.

The Exos (at least the X20, which I have) ignore the power saving options that you set in TrueNAS. Unlike using APM values, they instead use EPC. You will need to use Seagate’s SeaChest Utilities to review and disable the EPC features.[1]

In my opinion, yes. The marginal benefit of parking the heads, which won’t stay parked long, is annoying at best, and could indeed add wear to the drive’s mechanical parts at worst.

  1. Preferably in a live Linux session, such as from an Ubuntu ISO. You don’t want anything in TrueNAS to inadvertently try to use or poll the HDDs. As “quiet” of an environment as possible. ↩︎

1 Like
5

Hmm. Any other benefits like having less errors, faster performance or just less hot etc?

Hmm. What does ashift actually define? The sector size of the drives in a pool?

OMG. I see now.

Hmm. I also think so cause going idle and then spinning up fast all of sudden when a user tries to access the pool/disk on a system, may put extra load as its a mechanical drive. I think it makes more sense to have the power savings feature on the SSDs.

6

Probably not. With an ashift of 12, the smallest write a drive will ever receive is 4-KiB anyways.

The unit of a pool’s (vdev’s, really) block size. The smallest possible write.

7

I see. So, should i keep it to 12 (default) or use 9?

Also, how can i check what ashift is being set/used on my TrueNAS server?

8

Leave it at the default.

There’s a reason TrueNAS defaults to 12. It doesn’t matter if your drives’ logical sector size is 512B or 4K.

1 Like
9

no hard drives manufactured today actually use 512 bytes per sector, they only emulate it (with the caveat that the OS should align partitions and consequently writes to it, as nearly all filesystems write at least 4k at a time anyways.)

as a result, ashift should at a minimum be 12 and this is what TrueNAS does, the only downside to a “too big” ashift is very slightly more waste in the filesystem (since the minimum sector allocation is bigger), which is better than potential performance problems of an ashift that is too small (and possible rejection of I/Os from drivers)

sudo zpool get ashift

10

Gotcha!

11

it seems my other post got lost in the wind…

you can absolutely convert modern drives using Seachest Lite or WD’s HUGO. it takes roughly 30 seconds.

if you do, you must make sure your HBA supports it (anything IT Mode LSI 9200 series and newer does) and expect to resilver the drive as this the existing data will be scrambled with the new geometry.

and also if you do, make sure you are on at least ElectricEel as there is a rare issue that can happen if you have 50+ 4kn disks, where I/Os could spuriously fail (can work around by limiting zfs_vdev_aggregation_limit, but just update)

I run all my disks 4kn, if you want to know anything let me know.

12

I’m using LSI 9400-16i and plan to use the 16*16TB Seagate EXOS X16 drives. I’ve already flashed the HBA card to the IT mode with the latest firmware version available on the Broadcom site. Currently, the disk has to go through firmware update, probably the sector to 4K and then burn in tests to ensure everything is fine with the disks and then start the deployment. So, as this isn’t in the production server, i guess i don’t have to do the replace and resilver thingy in my case.

Cool cool. The drives are not hotswap as its in a tower chassis. So, i guess shutting down the system and turning off the PSU and waiting for a few mins and then powering on the system will do the job, yeah?

Umm, Electric-Eel is for TrueNAS Core or Scale? I had a very weird bug recently with the TrueNAS Scale, where my 8 drives in other NAS all enterprise were giving USB like speed. I diagnosed everything, installed the OS multiple and no go. Then, i made a switch to Core and the speed was back again. Long time ago, when i was searching for this issue, i also found that it also happened to a fellow user on this forum in the past.

This time, for this big new server, i’m not sure what to go with either Core or Scale ;(

Any insights on this would be really helpful!

OMG. What server are you using? Is that with a JBOD chassis? Yes, but for whatever scenario, if 4K can do better, why not when there is an opportunity!

13

Hey ya, don’t worry. I was fast enough to reply to your old post. xD

1 Like
14

ElectricEel is the (at the time of writing) latest version of TrueNAS Scale, which is linux based.

this was a linux specific issue afaik, and it caused failed I/O (ie you’d see write errors in zpool status, you’d clear the pool and it’d resilver and instantly fix it) (when doing intense things like running a scrub, sending and receiving multiple datasets, etc on a very wide pool… not slow I/O so I don’t think your issue is related. but you can see the pull request that fixed it here and the issue that I ran into that told me how to fix it

Main server is a Dell Poweredge T630 18-Bay. most of the storage is in many custom build JBODs using the Adpatec AEC-82885T and LSI 9305-16e HBAs.

15

I see. Thank you for bringing my attention!

OMG. That’s so much disks. Dream setup :slight_smile:

16

@winnielinnie So, i was installing the drives and went ahead and thought to give one more read before i start to upgrade the firmware and change the sector to 4K.

While searching, i found this:

So, does this mean if we change the sector to 4K, the power saving mode is disabled or the timer for going in idle mode is ignored? I thought to check with you before i go ahead!

17

Couple of good sources of information on this:

18

You don’t just “disable EPC”. It will revert the next time you reboot.

What you do is disable the individual “saved timers” for the idle_X power settings. This will survive reboots.

This looks something like, assuming the same command name and the drive identified by the software is sg0:

SeaChest_PowerControl -d /dev/sg0 --idle_a disable
SeaChest_PowerControl -d /dev/sg0 --idle_b disable
SeaChest_PowerControl -d /dev/sg0 --idle_c disable

Ironically, you want to keep EPC enabled to be able to disable the timers. Go figure.

1 Like
19

It actually is depressing that we are still dealing with 512e drives ten years after this Dell document.
How long does it take for “the industry” to get rid of obsolete cruft?

3 Likes
20

We still had 32bit games and applications up until recently.

1 Like