SSD smart and scrub schedules?

[diskdiddler]

Do ssds even work properly for smart schedules?

Will smart detect all the right flags on ssds and what risk is there if any?

I will be finalising my system in the next few weeks and I want to be sure that I've configured it very nicely.

 

[Johnnie Black]

I don't run SMART tests on my SSDs since IMO there's not much point, but I receive a daily email with all the SMART attributes, as for the scrub schedule it as you would for any other pool.

 

[diskdiddler]

I don't run SMART tests on my SSDs since IMO there's not much point, but I receive a daily email with all the SMART attributes, as for the scrub schedule it as you would for any other pool.

So a long test is worthless, but short ok?

 

[Johnnie Black]

So a long test is worthless, but short ok?

No harm in running either, but IMO short is enough, I even had to disable SMART for my SSDs due to a bug with the Crucial MX500 firmware where it's constantly detecting one pending sector.

 

[jgreco]

No harm in running either, but IMO short is enough, I even had to disable SMART for my SSDs due to a bug with the Crucial MX500 firmware where it's constantly detecting one pending sector.

Is that actually a bug? Did you try repairing the sector?

It's unusual, but SSD's can have read failures too. When a ECC or checksum fails, the block is effectively errored in a manner similar to a HDD. The controller should remap upon a rewrite.

 

[Johnnie Black]

Is that actually a bug? Did you try repairing the sector?

Yes, I have twelve SSDs and they are all the same, I've also seen other users in this and other forums with the same issue, once in a while every SSD will report one pending sector, after a few minutes it goes back to zero, so every time I restarted FreeNAS I would get 12 warning emails in the first couple of days, one for each SSD.

 

[jgreco]

That almost sounds like there's a drive firmware issue that might be actually causing a true pending sector, perhaps due to lack of power-loss protection.

 

[Johnnie Black]

perhaps due to lack of power-loss protection.

Don't think so, I only mentioned rebooting because FreeNAS doesn't warn again after the first pending sector, i.e. it would only warn again if there were more than 1, but it keeps going from 0 to 1, back to 0, back to 1, etc, even without any reboots, you just don't get additional warning after the first one.

 

[jgreco]

Interesting. Strange.

 

[Johnny Fartpants]

Yeah I have similar behaviour with a Supermicro 36 bay using 36 x 1.6TB Intel SSD drives. Its been running for over a year now and during that time have had about 6 drives show 1 or even 2 pending sectors. Initially I swapped the drives and RMA them but after about 3 over a few months I stopped. Currently have 4 drives in the system that have complained about 1 or 2 pending sectors. On close inspected a couple have 1 reallocated sector and a couple don't so seems a bit random.

 

[Johnnie Black]

On close inspected a couple have 1 reallocated sector and a couple don't so seems a bit random.

That looks a different issue, the MX500 keeps cycling between 0 and 1 pending sectors, for all drives, sometimes they show up on the daily report, for example this is the report from Tuesday:

You can see 1 pending sector for ada6 at the time the report was made, this is the report for Wednesday:

Pending sector gone, and it keeps cycling like this, for all the SSDs.

 

[Johnny Fartpants]

Yeah sounds a bit different but Im not running any scheduled SMART checks on theses drives so perhaps thats why you're getting it on all your drives?

 

[jgreco]

Maybe a stupid question and you can tell me off if so: Have you guys checked for firmware updates for the SSD's?

 

[Johnnie Black]

Have you guys checked for firmware updates for the SSD's?

Yes, there was a new firmware released by Crucial, which apparently was pulled, since I can't find it now, that didn't fix it.

 

[Johnny Fartpants]

Its a fair point and mine certainly have an update available. Its a production system and Im not sure how keen I am updating the firmware on 36 drives at once

 

[diskdiddler]

So to bring this back on topic, to be clear, freenas is capable of understanding the smart short data, in a meaningful way which does genuinely indicate failures?

I know I saw a job on the bug tracker a while ago about starting to read this data better from flash drives. Don't recall the job #

 

[jgreco]

So to bring this back on topic, to be clear, freenas is capable of understanding the smart short data, in a meaningful way which does genuinely indicate failures?

I know I saw a job on the bug tracker a while ago about starting to read this data better from flash drives. Don't recall the job #

Last I checked, FreeNAS is basically reliant on SMART for this. Some of it is filtered by smartmontools, which maintains "fixes" for certain oddly-behaving devices. I'm not aware of any specific FreeNAS meta-awareness beyond what the entire SMART stack (including smartmontools) provides, but I could be wrong.

At the end of the day, SMART is the drive industry's preferred way to measure the state of disks, and if a drive is reporting false or stupid SMART data, this is really an issue for the drive manufacturer, not something that FreeNAS should have to work around.

Imagine if you were driving down the road and saw a sign that said "In our city: Yellow means stop. Red means slow down. Green means go."

 

[diskdiddler]

At the end of the day, SMART is the drive industry's preferred way to measure the state of disks, and if a drive is reporting false or stupid SMART data, this is really an issue for the drive manufacturer, not something that FreeNAS should have to work around.

Not sure if we're talking about the same thing here, or I'm misreading.
My point is, I THINK there's a couple of new smart values for flash, like "FLASHWEARLEVEL_RATING" for example - stuff like that, which is a new value, specific to SSDs over HDDs.
(I might be _entirely wrong_ but I'm pretty sure that's right?)

My question is, does FreeNAS, now identify, a drive is flash, not HDD and when doing so, it takes notes of these flash specific thresholds and report them (if applicable)

Answer is probably yes, but I thought I'd check the best practice for monitoring my new SSDs.
Thanks for reply!