Cyberjock moderator gone?

[Patrick M. Hausen]

Yes, yes, yes, yes, yes, there is.

Which is why I got to a point where I was putting out stickies for all the frequently asked questions, because answer quality declines as tediousness increases.

BTW ... I think two or three of my posts could be a "resource". How would I go about that?

 

[ornias]

BTW ... I think two or three of my posts could be a "resource". How would I go about that?

ctrl+C
click: https://www.ixsystems.com/community/resources/add
ctrl+V

 

[diskdiddler]

Yes, yes, yes, yes, yes, there is.

Which is why I got to a point where I was putting out stickies for all the frequently asked questions, because answer quality declines as tediousness increases.

I've done technical support in some form or another, for over 20 years now and it's hard work at times. Especially when people ask particularly silly questions. Gotta just grin and bare it, but boy!

 

[jgreco]

I've done technical support in some form or another, for over 20 years now and it's hard work at times. Especially when people ask particularly silly questions. Gotta just grin and bare it, but boy!

It's even worse for ZFS, where people come in with preconceived notions, or worse yet the "I been done an IT guy for twenty years, there ain't nothin' I don't already know" attitude (the former is fixable and I'm happy to try, the latter is usually clue-resistant and can be unpleasant to impart clues upon).

Grin and _bare_ it? Okee...

 

[ornias]

It's even worse for ZFS, where people come in with preconceived notions..

To be fair, Cyberjock clearly was quite guilty at that too himself at times ;)

 

[Redcoat]

One of the attractions for me to FreeNAS and this forum "back in the day" was the fact that FUD and poor advice/info posted in the forum was most likely going to be identified, challenged and/or debated - albeit often in a most aggressive and potentially off-putting way to many readers - but not just ignored.

So I judged this environment overall likely to result in a higher-quality information source than others I had observed, thus my confidence in it.

In their own ways Cyberjock and jgreco both were noted contributors to that confidence.

 

[ornias]

overall likely to result in a higher-quality information source than others I had observed

Overconfidence and agressive environment leads to hard rooted misconceptions getting mixed in. Like the "1gb ram per tb" rule/guideline, which only keeps getting used by people because of this forum in the old days and it agressively being presented as the truth.

often in a most aggressive and potentially off-putting way to many readers

This leads to a situations where a lot of people who know things are factually incorrect or misunderstood, just skipped the forum. Not just readers. If I come on a forum with people fiercely defend bullshit, I just close the tab.

 

[jgreco]

Overconfidence and agressive environment leads to hard rooted misconceptions getting mixed in. Like the "1gb ram per tb" rule/guideline, which only keeps getting used by people because of this forum in the old days and it agressively being presented as the truth.

I basically wrote that into the manual. The reason that this was presented as the "truth" was because we were seeing people come in with fatal pool corruptions when they had seriously mis-sized systems, and it was deemed important that this stop. It's extremely difficult to get precise debugging out of end user systems, and iXsystems wasn't interested in throwing developer resources at it to replicate the issue, because it was far outside their use case as well. Remember, FreeNAS is basically TrueNAS-ALPHA.

The worst were the AMD APU's trying to come over from FreeNAS 0.7 and use ZFS. 4GB systems were a bad thing. 8GB systems seemed to be fine as long as there wasn't a stupidly mismatched amount of disk.

So, having watched a bunch of cases of this, I noticed some trends, and then I identified some RAM sizing at which this didn't seem to be causing problems. I specifically wrote the 1GB RAM per TB to be a bit vague. It never said, for example, whether that was per raw space, available space, or used space. The basic goal was simply not to have stupidly far off amounts of RAM, because this had been observed to cause problems. In general, ZFS needs RAM to be dimensioned larger as space goes up because the amount of metadata, metaslab free info, etc., goes up, and performance suffers when there's insufficient resources. The specifics may be different for varying use cases, but the fundamental truth is that RAM is good, don't cheap out on RAM. And if you don't give people some sort of rule of thumb, then they often try to "argue you down" as though this is just some arbitrary thing you're making up, rather than a compsci design tradeoff, which is actually what it is.

You can say that we were presenting something that wasn't necessarily "true" as a truth, because, in fact, many people did manage to run ZFS on 6GB and got away with it. But there was risk in running on low RAM, and those of us doing volunteer support work on the forum wanted users to run a safe configuration.

So, @ornias -- should I have done something different, and if so, what?

 

[Arwen]

One point I do want to make, is that FreeNAS is not for everyone.

We sometimes get people here who insist that their NAS must run with one or more of the following;
- less that the required 8GBs of memory
- must use USB attached data disks
- must support SATA port multipliers
- must be super Windows friendly, (almost to the point it needs TO BE Windows)
- must spin down disks
- must support large RAID groups, (aka >12 disks in a single ZFS vDev)
- must support and be reliable, non-RAIDed data disks
- must run on a really low end computer, like Raspberry Pi

Some of these are recipes for disaster, others less than ideal. I calmly state what I know about their particular selection. And if they still insist that FreeNAS must do it the way they want, then I suggest they look elsewhere.

I remember a guy wanting a NAS without protected disks, (no RAID-Zx or Mirroring). He said he had ways of getting the data back. But then he said he wanted a large pool, like 10 or more disks. Fine, except if you loose 1 disk, you loose the whole pool. He kept insisting that FreeNAS should not work that way. He only wanted to restore the affected disk. I gave him 2 choices, one pool per disk or use some time of RAID-Zx. Neither is what he wanted. Don't remember where it ended up.

 

[jgreco]

Some of these are recipes for disaster, others less than ideal. I calmly state what know about their particular selection. And if they still insist that FreeNAS must do it the way they want, then I suggest they look elsewhere.

Yeah. I do try to make sure people hear the things they need to know, and I may be pushy if they don't seem to comprehend them. But if you are willing to own your own mess, yay, everybody deserves to be able to do as they please.

 

[ornias]

I'm not really into debating this. IMHO if you go cherrypicking examples instead of the general point someone tries to make, it isn't worth discussing it any further.

So, @ornias -- should I have done something different, and if so, what?

That being said I do want to answer your question somewhat without going too deep into the subject at hand:

Minimum 8gb, recommended 16gb, optimal 32gb

Make it seem like "just another piece of software", Now it seems like there is some kind of ZFS magic. (well there is, but I think the users asking about the precise way of how caching and metadata sizes can be calculated isn't your target audience)
Which (not totally coincidental) aligns somewhat with the IX lineup ;)

KISS and "Don't feed the Troll" basically:
Simple enough for everyone to grasp and doesn't open you up to endless discussions about whether your rule of thumb is correct or not (which is also the endless discussion I tried to avoid with this reply that took me the better part of an hour,)

 

[ornias]

...
Some of these are recipes for disaster, others less than ideal. I calmly state what I know about their particular selection. And if they still insist that FreeNAS must do it the way they want, then I suggest they look elsewhere.
...

This, some people only want to hear their own idea fed back to them. No rule of thumb or recomendation would help against those. "Take it or leave it" is really the only solution, because often they are also not really interested in actually understanding why something won't work. Which is the most annoying type to deal with in IT.

Don't remember where it ended up.

The data? Thats gone for sure. ;)

 

[Arwen]

This, some people only want to hear their own idea fed back to them. No rule of thumb or recomendation would help against those. "Take it or leave it" is really the only solution, because often they are also not really interested in actually understanding why something won't work. Which is the most annoying type to deal with in IT.
...

Yes, at some point, you have to "Stop feeding the Trolls". Even if the person is not a "troll", their actions can appear to be. And at some point, I am not not going to, well, waste my time on explanations.

Today though, our forums have:
- Useful stickies
- Resource section
- Calm and clear responders

Some of the things that come up more or less regularly, get either a sticky forum thread. Or a resource. For example, I started using an external disk for my NAS' backups. The subject came up often enough that I wrote a forum resource on the subject. Anytime someone asks that particular question, if they have not been responded to already, I point them to my resource.

 

[Apollo]

Minimum 8gb, recommended 16gb, optimal 32gb

I would emphasize on proper casing for units.

Minimum 8GB, recommended 16GB, optimal 32GB

 

[ornias]

I would emphasize on proper casing for units.

Minimum 8GB, recommended 16GB, optimal 32GB

I care as much as for the grammar being utterly fucked up.
Its the internet, not a paper.

 

[danb35]

Given that the question's been thoroughly answered, and subsequent posts are either complaints about someone who's long since gone, or only (at best) tangentially related to the question posed, might I gently suggest it's time this thread be closed?

 

[jgreco]

I'm not really into debating this. IMHO if you go cherrypicking examples instead of the general point someone tries to make, it isn't worth discussing it any further.

How is talking about the exact thing that YOU called out, "cherrypicking"? Especially when I am the party responsible for enshrining that. **confused**

That being said I do want to answer your question somewhat without going too deep into the subject at hand:

KISS and "Don't feed the Troll" basically:
Simple enough for everyone to grasp and doesn't open you up to endless discussions about whether your rule of thumb is correct or not (which is also the endless discussion I tried to avoid with this reply that took me the better part of an hour,)

But that's the thing. The GB-per-TB guidance was the simplest thing that generally fit and was easily understandable. It wasn't possible to say "8GB minimum, 32GB optimal" for LARGE pools, because that absolutely does NOT work.

It seems to me like you're telling me that the rule of thumb was bad, but then on the flip side, come up with an easily understood thing that works... and that was the rule of thumb. Well, actually "8GB minimum, 1GB per TB of space." I don't think we've seen catastrophic problems with that, unless you try to do dedup, which is not recommended anyways.

 

[jgreco]

Given that the question's been thoroughly answered, and subsequent posts are either complaints about someone who's long since gone, or only (at best) tangentially related to the question posed, might I gently suggest it's time this thread be closed?

Yeah, fine, this isn't really going to ever be productive.

 

[Ericloewe]

BTW ... I think two or three of my posts could be a "resource". How would I go about that?

PM me for guidance.