If the author meant ashift, he would’ve written ashift.
But he didn’t.
This is what we call, confirmation bias.
You are literally reading a word that isn’t there in order to make your argument work for you.
That’s how wrong your argument is, that you literally have to make stuff up that doesn’t exist, just to make your argument work out for you, rather than reading what’s actually there, and still trying to find a way to make your argument work for you inspite of the word that’s actually there.
Because if you’re just going to make shit up that isn’t actually on the page, two can play that game as well.
You want to replace sector with ashift. However, using your own logic against you, I can, ad simile, replace sector with recordsize and then you’d be equally screwed.
It’s stupid to argue about things using stuff that isn’t there, because then you can make it whatever the hell you want.
Jesus christ, yes!
That is what I wrote! It can be anything!
But we use 16 sectors after compression for our example. Because that is the example the author uses.
No. Like already explained, if the author would have used ashift, he would have needed two different spreadsheets. Because it does not behave the same.
I really think he is just pretty stubborn. If he would manage to put away his pride for a moment and play along with the examples, he would realize it pretty quickly.
Because you can’t actually make the spreadsheet, which doesn’t say what you want it to say, for it work out for you, so that your argument would work???
LOL…
Really???
The best that you can do, in an argument, when the argument doesn’t work out in your favour, is to call the person whom you’re arguing against, brain damaged?
You are, in your example, literally and explicitly is still referring to
That is literally what you wrote.
Funny how now, all of a sudden, you’re actually abiding by what the author of the spreadsheet wrote. Interesting, eh? You weren’t/still aren’t abiding by what the author of the spreadsheet actually wrote in terms of sectors, as you still think that I am “confused” (remember, ever accusation is a confession when it comes from you).
Compression can play a role in it, but it doesn’t have to.
If you create the ZFS pool using the command zpool create, compression can off by default, depending on which version of OpenZFS you’re using and/or when you created the pool.
From the drive/ZFS spacemap geometry perspective, the only relevance that compression will have on this is how many sectors it will actually take, to commit your data to stable storage.
Thus, if you have 16k worth of zeros, the actual amount of data that would be committed to stable storage would be relatively negligible.
So, if you want to abide by the author’s example of using compressed data, expressed in the number of sectors, there are scenarios/situations where talking about how many sectors data takes, when committed to stable storage, can actually work against your own example (because zeros and text files generally have a high compression ratio, and therefore; would work against the point that you were trying to make, with your example).
So if you want to abide by the author’s example, I’m okay with that. Just know that it is may not conform to the point that you were trying to make, with your example.
^See you gaslighting me above.
Again, you are the one who is making the argument/claim that you are not abiding by what the author actually wrote (sectors) and that you want to read and interpret “sectors” into making it whatever you want it to be.
And if that is how you want to engage in your quote “bad faith” argument (again, every accusation you make is a confession), then likewise, I can make sector mean anything that I want it to mean as well, ad simile, and I choose to make it mean exactly what it literally says: sectors.
Two can play that game if you’re just going to be gaslighting me/making stuff up anyways.
It is always interesting to me, that when people realise that they’re losing an argument, that they’ll stop arguing about what’s actually on the page, and start making stuff up to argue/talk about instead.
Neither of you can’t stop gaslighting me into thinking that sectors doesn’t actually say sectors.
Like how pathetic is it, for both of you, that you’re still arguing with me about what the author of the spreadsheet wrote, when all I had to do was to cite the spreadsheet back to you guys, and you still want to argue with me about what the author actually wrote, rather than to argue with the author of the spreadsheet himself?
Why?
Cuz y’all are chickens. That’s why.
You’d much rather argue with me (the messenger) about what I am throwing back, in to your faces, because it’s “easier” for you to do that, than it would be for you to actually take up the argument with the actual author of the spreadsheet himself.
That’s not arguing from a position of strength. That’s arguing cuz both of you are chickens.
As I have said before, if you have or take issue with the fact that the author of the spreadsheet used the word “sectors” (TWICE per tab, in the entire workbook, no less), take it up with the author.
But y’all are too chicken to do so, and this is why you are continue to argue with me about something that I’m not even the author nor responsible for/of.
Take it up with him.
But you’re not going to.
Why?
Cuz both of you, are chickens.
This is why we’re still arguing about this cuz y’all are chickens.
You know that the right thing to do here would be to find the email address of the author, and reach out to him, to ask him about it.
But neither of you have actually done that, and instead, are arguing with me, and employing ad hominem attacks rather than learning how ZFS works with sectors.
George Wilson talks about it, in terms of drive sectors. Kirk McKusick talks about it, in terms of drive sectors.
You want to continue this argument with me where I have the sources and citations from Matt Ahrens, George Wilson, and Kirk McKusick?
Really???
Like do you not realise how much of a losing proposition that would be for you to continue this argument with me about how sectors doesn’t mean sectors when literally the guys that created ZFS talks about spacemap allocation, in terms of sectors.
37:43
controls a lot of the way that we do allocations we can then turn around and say okay well you know we can give you
37:49
variable block size because we know not every block size fits all workloads so if you have a bunch of tiny files why
37:56
would I give you an 8k block so that you could store 512 bytes in it when I can just simply give you 512 bytes and it
38:01
stores uniquely your data so we have this concept of variable block sizes that are powers of 2 all the way up to
38:09
default 128 K now we actually have large block support we go up to 16 Meg 1 Meg
Like I said, you’re more than welcome to take up your argument with Matt Ahrens, George Wilson, and Kirk McKusick. Good friggin’ luck with that one.
The more important question is that you finally agree that he’s talking about sectors, notashift.
Do you agree with that?
I’m not answering the question until you can demonstrate that you understand that drive sectors came wayyyy before alignment shift, and that you also, demonstrate, that you understand what “alignment shift” is, and what it does, and what it means.
Once you can positively confirm that you have finally understood what alignment shift is, what it does, and what it means, then and only then will I answer your question.
You want an answer to your question.
Positively show that you understand what alignment shift is/means/does and then we can continue from there.
Why must you engage in gaslighting where you are telling me that what I am literally seeing/reading, is not what I am literally seeing/reading?
That’s the definition of gaslighting.
Why must you employ gaslighting, in your argument?
More gaslighting.
Cuz you’d literally rather gaslight and argue in bad faith because you’re not arguing to actually learn how ZFS actually works.
Again, every accusation is a confession.
And yet, you still can’t prove anything that I have said/wrote is wrong, without you resorting to gaslighting.
The only way that you view your ability to win this argument is if you literally gaslight me.
That’s the only way that you can view yourself winning this argument.
i.e. you can’t win this argument on its own merits. This is why you have to gaslight (me) (so you think you can win this argument).
That’s the only way that you’re able to make this argument work for you.
Not on its merits, but by literallygaslighting.
if you were so confident that I was wrong, the would’ve email the author of the spreadsheet by now and asked him if by “sectors” he really meant “ashift”.
But you haven’t.
And therefore; you don’t have the reply from said author of said spreadsheet either confirming or denying your argument.
If you were so confident about your argument, you would’ve already gotten that email from him, pre-emptively, but you didn’t.
Yes. Exactly.
He is talking about sectors, which either are 512 or 4k (in 99% of the cases), and because of that 16 sectors are either 8KB or 64KB, according to him. Do you agree with that?
Sure. Because by answering it, you might even realize where you are wrong.
ashift is, and that is the very important and critical part, is a setting of your pool. You can set it to whatever you want. So it isn’t connected to any underlying hardware directly. You can set ashift to 13 or 8k, even if your pool hardware uses 4k.
Drive sectors are what you actually use in hardware. That would be by a 99% chance 4k. Like you already demonstrated, besides some niche and obscure, overprices and small 4TB HDD, 512 are mostly nonexistent anymore. Sure, you will finde maybe some legacy Oracle DB.
While SSDs sometimes use something bigger than 4k, most of them have the controller tuned to 4k, so it is save to assume that 99% of all drives use 4k.
Good thing is, you don’t have to care. By default ZFS uses ashift=0 which means auto detect.
So since we know now that 99% of the drives come with 4k, the smallest sector such a drive can write is 4k. For a 512 drive it would be 512.
In general it is of course best to match ZFS sector size (ashift) with the drives sector size.
And in general that happens automatically, because like already said ashift=0 is the default.
A smaller ashift offers a better pool geometry, so you would want to set it as low as possible from a ZFS standpoint, but not lower than what the drive actually is capable of from a hardware standpoint. This leads to the very simply rule: match ashift with what your drive is actually using. Which, again, happens by default.
Just for fun, we can go trough what happens if ashift is wrong.
Assuming you set the ashift to low:
I don’t know what would happen if you set ashift=9 on a 4k drive. Probably it would just need to allocate 4k for every 512 sector?
Assuming you set the ashift to high:
If you set ashift=13 on a 4k drive, you won’t gain anything. Drive simply has to write two sectors on the hard drive. Chances of a good pool geometry is now worse, because of that storage efficiency is worse.
But since this stuff all gets correctly auto detected anyway, it is mood to discuss.
No rule without exception Micron 7450 server SSDs had a firmware bug, that lead the drives to tell the system they are using 256k sectors. Of course that was not true and they use 4k like everybody else. They fixed it with a firmware upgrade. https://www.mouser.com/PCN/Micron_Technology_PCN_34996_Micron_7450_E2MU200_Change_List_External.pdf
Money quote: “Changed Namespace Preferred Write Granularity to 4KB”
So, since I answered your question, would you please now answer mine?
He is talking about sectors (like you said), which either are 512 or 4k (in 99% of the cases), and because of that, 16 sectors are either 8KB or 64KB, according to him. Do you agree with that?
Did not have the time yet, or realized you were wrong but too proud to admit?
It is totally fine if you write another wall of text, but please don’t forget to answer the yes or no question:
He is talking about sectors (like you said), which either are 512 or 4k (in 99% of the cases), and because of that, 16 sectors are either 8KB or 64KB, according to him. Do you agree with that?
99.999999% of the population that uses drives of ANY kind, aren’t going to bother doing that.
Therefore; whereas ashift, as you stated, can be a user defined variable, sector sizes, is not a user defined variable.
That makes the calculation, on the spreadsheet, far more stable.
Like, how do you stillnot get this?
Bullshit.
There are still PLENTY of drives that uses 512b sectors.
Companies like Backblaze still use PLENTY of drives that are using 512b sectors.
You literally write stuff from your ass, that has no basis in reality, whatsoever, when it’s so easy for me to find data sources that disprove the statements that you pulled from your ass.
Oracle DB isn’t a drive.
It’s an application.
Why do you need the idiot in the room to explain this to you?
No! Patently false.
You cannot assume that 99% of all drives use 4k.
Again, you’re just literally pull this shit outta your ass.
If the drive reports the sector size to the system that’s hosting ZFS, correctly.
Again, the OpenZFS documentation is explicitly clear about why alignment shift is a thing they introduced to begin with.
Read it again.
Why do you have to pull stuff outta your ass when the OpenZFS documentation explictly tells you what it is and why it’s there?
Note your claim here:
And yet, you spent most of your reply, talking about drive sector sizes, and notashift, which is what I’ve been pointing this out to you, a week ago.
Again, if you were actually interested in increasing your understanding of how ZFS works, you would’ve acknowledged that you were to wrong to equate “sectors” as “ashift”, but even after spending this comment, talking about drive sectors, you still admit that you were wrong for making that incorrect equation.
How do I not get this? I don’t even disagree with it
The person making the claim that Oracle DB is a HDD, is he in the room with us right now?
I give up. You get out of your way to misunderstand me (we all know that I was talking about software uses cases for 512 and not “Oracle HDDs”), don’t engage with what I wrote or asked, and are to stubborn to learn anything, this leads to nowhere.
It was fun, but now we are going in circles.
Yes! You! You’re the one who is making that claim.
This is what you literally wrote:
You were literally talk about 512b sector hard disk drives, not the software, application, nor the data that’s stored on said hard disk drive.
If you express yourself poorly, that is not the responsibility of the receiver to have to decipher what the fuck you were trying to talk about, but rather it is your responsibility to express yourself exactly as what you intend to say.
(If you don’t say exactly what you mean and you leave it up to the receiver to interpret what you have said, and more importantly, what you didn’t say, then what you didn’t could be what the fuck you want it to be. I’m not a fucking mind reader.)
You sound like my wife, when we first got married.
Then what’s the fucking point of you writing anything???
LOL…LMAO…
Really???
It literally took you a fucking week to learn that the spreadsheet was literally expressed in drive sectors, not ashift, per you:
You state that you don’t even disagree with it, and yet, you have, literally spend the entire fucking week, disagreeing with it.
Say the person who spent the entire fucking week, arguing about something that you ended up agreeing with my original argument, that I put forward, a week ago.
And yet, you spent said entire fucking week, arguing about that very thing that yourefused to learn about, even though you could’ve literallyYouTubed that shit since the start of the argument, but you didn’t (for a variety of reasons).
I’ve been telling you about this, for a week.
You were the one, who refused to learn, and then spent the week, arguing with me about it.
Micron 7450 server SSDs had a firmware bug, that lead the drives to tell the system they are using 256k sectors. Of course that was not true and they use 4k like everybody else. They fixed it with a firmware upgrade.