WD RED - Western Digital idle3 - What the Heck? - Which setting is the best? - Some WD BLUE talk too

[webdawg]

So I use pfSense and searching into timeout errors where FreeBSD ejects a blue western digital laptop drive I have came across the wdidle3 util and the wd5741 designed specifically for the WD Red Drives. I have figured out (from reading) that the wd5741 util sets the (on the red drives) Intellipark feature to 300 seconds.

What is the deal with this? I have blue drives in laptops that are up to 1302656 parks that were set to something like 4 seconds. How is this supposed to help at all? Does it save the head electronics? I get that if the drive is in the laptop it could save it during a drop or something.

But here is the bigger question? Why are they parking WD Red drives? I still have yet to look at my array of 4TB reds at home but I am sure they are parking after 300 seconds.

I had to disable parking on my pfSense (at least I think, another month+ to test) so it would stop detaching the WD blue drive ( I do have other solutions in mind).

Back in the old days you would HAVE to park a drive head before shutdown/movement of the drive, but what does parking after 300 seconds do for me or the drive?

Also:

Why does hdparm have this in its man page:

A setting of 30 seconds is recommended for Linux use.
Permitted values are from 8 to 12 seconds, and from 30 to 300
seconds in 30-second increments. Specify a value of zero (0)
to disable the WD idle3 timer completely (NOT RECOMMENDED!).

Why is it "NOT RECOMMENDED!"

I get that setting to 0 on the WD--EZRX drives seemed like it would park the head just < then 1 second heh but on drives that it works like the blue drives that I am talking about, what does it matter? If I need the head never to park, what is wrong with that?

 

[nightshade00013]

The aggressive head parking allows WD to say, "Hey our drives use less power which saves YOU money." the problem being that every time the heads park they wear more and eventually end up costing more money that the electricity since you have to replace it and probably lose data and a lot of time trying to fix a problem. More electricity used also means more heat generated.

You probably want to park the heads if it has been idle for five minutes since the likelihood of having a write occur JUST after the five minute mark is rare. Generally if a drive has been idle for that long the person has walked away and will be away for a while longer. Setting it to zero does create extra power use and heat but you have to weigh your use case scenario and decide what works best for your situation.

WD is also trying to get to a point that a drive can be fine tuned for a particular situation, it means they have more products coming off the same line that only require a slightly different firmware for what can be a vast difference in price. Why sell one drive that can do all the jobs when you can make multiple drives and charge a higher price for one over another.

 

[webdawg]

Except, I have a script that writes a file every 5 mins...so you are saying it is rare but it happens, it does happen. I cannot even remember how many cronjobs that I have and I am sure the log files write and write and write.

I just did a smartctl -a on a drive that has been in a router for the past 4 years and we have over 2 million parks here: 2886367

 

[nightshade00013]

If the script is based on CRON it probably runs every 5mins approximately, The CRON system has triggers and they happen at certain times. So if your script is set to run every five minutes it could be anywhere from 5 mins to 5 minutes 59 seconds. Those extra seconds can bite you in the ass when the heads park and then unpark. I can remember the exact way cron does this but I remember reading some information about it.

But yes it can and will happen especially if the script writes a very short amount and then takes a little long to run it's next iteration.

 

[webdawg]

I have my red drives in a fully cooled server rack.

Where did they expect people to put these things at, I mean even consumer NAS units use fans?

How much heat and power is the drive head going to use? It cannot be much...its a drive head.

They wear the hell out of the head just to save this little bit of power?


I guess the only question I need answered now is why does hdparm reccomend NOT disabling head parking?

 

[webdawg]

If the script is based on CRON it probably runs every 5mins approximately, The CRON system has triggers and they happen at certain times. So if your script is set to run every five minutes it could be anywhere from 5 mins to 5 minutes 59 seconds. Those extra seconds can bite you in the ass when the heads park and then unpark. I can remember the exact way cron does this but I remember reading some information about it.

But yes it can and will happen especially if the script writes a very short amount and then takes a little long to run it's next iteration.

I hear you? Isn't that great? So it could be every 5 mins for the past few months my drives hav been parking and unparking.

 

[Mirfster]

I have my red drives in a fully cooled server rack.

Pics or it didn't happen (really I am just curious)... ;)

 

[Ericloewe]

I have my red drives in a fully cooled server rack.

Where did they expect people to put these things at, I mean even consumer NAS units use fans?

How much heat and power is the drive head going to use? It cannot be much...its a drive head.

They wear the hell out of the head just to save this little bit of power?


I guess the only question I need answered now is why does hdparm reccomend NOT disabling head parking?

The added drag is surprising. And it's impossible to avoid, because you need the heads to generate lift, which induces drag.

 

[webdawg]

Pics or it didn't happen (really I am just curious)... ;)

Here:

 

[webdawg]

The added drag is surprising. And it's impossible to avoid, because you need the heads to generate lift, which induces drag.

What do you mean? Are you still referring to how much power it takes or?

 

[Mirfster]

Hey, I see your "Server Room" is about the same dimensions as mine. :)

 

[Ericloewe]

What do you mean? Are you still referring to how much power it takes or?

While the head is in an I/O position, as opposed to being parked, it's disturbing the airflow, causing additional drag. The drive's motor needs to compensate the added drag to keep the drive spinning at the proper angular velocity.

 

[webdawg]

While the head is in an I/O position, as opposed to being parked, it's disturbing the airflow, causing additional drag. The drive's motor needs to compensate the added drag to keep the drive spinning at the proper angular velocity.

So is there any additional wear on the motor in this case?

 

[Ericloewe]

So is there any additional wear on the motor in this case?

Well, that's had to say. I'd bet on negligible added wear on the motor. But it does have to work harder, which means more power.

 

[webdawg]

So besides the power and heat, anyone see any other reasons not to disable head parking?

 

[Ericloewe]

So besides the power and heat, anyone see any other reasons not to disable head parking?

Oh, don't misunderstand, definitely disable it if your workload is not appropriate for it. I'm just explaining why they bother parking the heads.

Though mine aren't set not to park, they effectively never park because of .system dataset writes.

 

[Bidule0hm]

So besides the power and heat, anyone see any other reasons not to disable head parking?

It also means you can't have head crashs when the heads are parked which is always a good thing to have I guess :)

 

[Ericloewe]

It also means you can't have head crashs when the heads are parked which is always a good thing to have I guess :)

Yeah, though I don't expect that to be a big problem on a server...

 

[nightshade00013]

Yeah, though I don't expect that to be a big problem on a server...

With some of the earthquakes I have felt here in Oklahoma over the last two years I could see it happening. One was basically like being on an elevator that slipped a little and dropped 6 inches. If the heads were flying at that point I don't want to know what might have happened. If you live near an active fault it could be beneficial as well at least as long as you are not having the .system data writes on the disks.

 

[Ericloewe]

With some of the earthquakes I have felt here in Oklahoma over the last two years I could see it happening. One was basically like being on an elevator that slipped a little and dropped 6 inches. If the heads were flying at that point I don't want to know what might have happened. If you live near an active fault it could be beneficial as well at least as long as you are not having the .system data writes on the disks.

Well, we've had an earthquake that literally reshaped thought in Europe in its aftermath, so I guess that qualifies as "near an active fault".