Build - Budget 96TB (Raw) Plex Multimedia NAS on Linux with ZFS

[@nekollama3994]

I recommend everyone to build a server like this. Really easy way to learn linux, and plex solved all my media problems that is was having.

[@AmmarossPlays]

Yep :D

[@peterjansen4826]

Here a little secret: Linux actually is incredibly easy. :p Seriously, installing and using Linux is probably easier than installing and using Windows. You just need to learn a few new tricks, that is all. (Arch and dwm for me)

[@linuxpirate]

Jellyfin is the way to go if you want an Open Source media server! (Fork of Emby)

[@zocker1600]

@@linuxpirate this^^, was about to say that Emby went closed source despite them claiming to be an "open media solution".

[@Aturnadagar]

Linux have come long way, but yeah, is way easy now.

[@LethalBB]

Wait a minute, you're not Scottish......

[@saintuk70]

And just to confirm, he's def not Scottish.... I can tell.... I am Scottish - lol

[@AmmarossPlays]

Nope. One of the writers :)

[@RobBCactive]

Thanks for the new content, still sorely miss Jim's work

[@AmmarossPlays]

Thanks!

[@PlasmaJunkie]

Whatever happened to Jim’s work? Is he just so busy with running the website that he doesn’t create this content anymore?

[@theDaNi0]

@@PlasmaJunkie Check out the "You're being misled" video. Jim threw in the towel when it comes to tech, he'll be refocusing onto more science stuff and let others do the tech (if I recall correctly).

[@introvertplays6162]

This is the final realization that there will never be a video from Jim ever again on this channel.

[@Brodda-Syd]

Not the usual in-depth documentary but as a Linux user I enjoyed this anyway.

[@phil.4688]

Friendly reminder that, apparently from what I read from e.g. ServeTheHome and Backblaze, there is non-trivial likelihood of mechanical drive failure upon re-silvering (it goes up to 7% average risk on typical WD/Seagate/Toshiba/etc drives). That's why Raid 6 / Z2 is effectively recommended for mechanical arrays - because 1 down has a 7% chance to become 2 down upon re-silvering, which would thus fail at that point in Raid 5/Z1. I'm telling you this because it's important to keep as many spares as your parity number on-site (so Z1 = keep 1 spare, Z2 = keep 2 spares, etc.) Also consider hot-plugging those as "spare" declared as such in ZFS, which will automatically re-silver upon failure). Because when your Z2 is essentially degraded to Z1 upon failure of 1 drive, you really want to have that second drive handy ASAP if you're going to need it. Statistics again say that you'll need that second drive 7% of the time after a first failure. You'll lose any parity on the Z2 array when that happens - so not only should the resilvering happen ASAP at that point, make sure you have a backup ready just in case!

[@AmmarossPlays]

Let me clarify a bit. You have a high likelihood of having additional errors during rebuild. Traditional hardware RAIDs will eject drives for minor errors like retries/bad bits/etc, but ZFS is more resilient and will be able to gracefully correct bad block data with checksums and/or parity data. So, even with raidz1, if you're rebuilding a lost drive and you hit a URE or bad bit, it can still recover, or at the very least have a single file corruption rather than losing the array. STH/Backblaze data were likely based on traditional hardware RAID. That said, RAID5 might as well not exist when you're talking RAID with disks larger than 2-3TB, especially with the 8TB we're using in the vid. Just the statistical likelihood of hitting a URE is nearly guaranteed at 6TB.

[@GeoStreber]

I just wanna point out, that you don't need Putty anymore under Windows. It now supports ssh by default. Go into cmd or powershell. The command is "ssh username@address"

[@AmmarossPlays]

True. I still use putty for a lot of the admin stuff though due to extra features, so force of habit to just bring it in here rather than using something built-in like Powershell.

[@klobiforpresident2254]

"Only two terabytes." My 60G hard drive took that personally!

[@klobiforpresident2254]

@SolidSnakeOil It's just the old drive in the PC connected to the printer.

[@AmmarossPlays]

Yep. That old system had eight of those small drives ;)

[@Freethinker]

Superb job Kirk!

[@AmmarossPlays]

Thanks Jim!

[@emdea]

running OMV on an old fujitsu home server. It initially had a celeron single core and 512mb ram but i upgraded it using old parts i had sitting around to a dual core and 2gb and now its running cloud sync, backups, media server (DLNA+bubbleupnp) teamspeak and some other monitoring stuff. It was a great learning experience to set it up and its proven to be quite useful so i highly recommend it to anyone who isn't afraid to tinker and learn. Good work on the video btw :)

[@peterjansen4826]

I did not expect to see any Linux-content from AdoredTV. Linux also is great for regular desktop-usage, you do have to make some sacrifices for gaming (maybe 20% of your game-collection) but there also are substantial advantages.

[@AmmarossPlays]

The advantage of having the writers help make content; we all have different skills. :)

[@churchseraphim1380]

The video quality is outstanding. Feels like I'm watching a much bigger channel than I actually am! Interested to see more content like this from the whole team

[@AmmarossPlays]

Thanks! Pulled out a few new tricks for this one.

[@martineyles]

Confused as to who you are. Have now had a look through and see you were in Overvolted, which I never got round to watching. Does this mean that the channel has been passed on to you, or to all the people from Overvolted? Could you do a channel update/trailer video where you explain who is posting and what content we should expect. I subscribed to the channel for Jim's analysis, and was disappointed that he got hounded out by others in the community, as he was doing a good job. Should I expect any of this type of analysis on the channel, or will it be more builds or just one offs. Is Overvolted home to some of the analysis content?

[@DanielLearmouth]

As I understand it, Jim has moved on from doing tech videos, and thus passed the channel onto the Overvolted guys. Jim moved onto a new channel called Freethinker, which appears to be about his other interests, primarily surrounding science and discussing and debunking certain theories surrounding it.

[@DonnyStanley]

@@DanielLearmouth This is correct.

[@martineyles]

Thanks, found Jim's new channel. Still, I would be interested to hear what the new plans are for this channel. Will the Overvolted guys do analysis or do they have other plans.

[@AmmarossPlays]

Jim has started a new channel called Freethinker that he's devoting a lot of his time to. The writers from the AdoredTV website, that's been running for a couple years now, will be making video-form content for this channel. Jim was working on an explanation video to post here as well.

[@djamrit510]

Hi, did you configure the boot drive in a RAID configuration? Does the LSI 9211-8i allow to create multiple RAIDS, e.g RAID1 for OS, and the RAID5 or RAID6 for storage? Thanks!

[@blackmennewstyle]

Server case, ton of hard-drives, GNU/Linux, ZFS, how can people dislike a video like that :( Happy midweek and stay safe during the human malware pandemic outbreak

[@peterjansen4826]

Probably gamers who are not interested in this content. :)

[@TristanKingly]

For PLEX it makes a lot more sense to go intel, just because of the free iGPU transcoding you get with quicksync. My i3-9100 unraid nas can handle up to 15 1080p transcodes. Hate intel but you can't beat that transcode performance for the price.

[@AmmarossPlays]

Yep. I explain that exact point in the companion article on the website. I couldn't talk them into pumping the extra $100 for an i5 though (oddly enough), as it was meant to do more than just Plex.

[@VADemon]

but for 80 TB raw and ZFS you want ECC, which is free for AMD-based platforms.

[@justacomment1657]

@@VADemon zfs does not trust ram nor the drives. So you can get away without ecc memory. I too would have chosen Intel for the igpu or a different media server emby for instance... I don't know if it can use the igpu of either to transcode but at least it can run on bsd... So you can get rid of those pretty unstable and old openzfs clones..

[@VADemon]

​@@justacomment1657 > ZFS does not trust RAM How is that *even* supposed to work? It can either trust drive data (why then keep metadata in RAM?) or RAM metadata (ECC please); double-checking RAM metadata brings us back to the former. OpenZFS documentation states: "Bit flips can have fairly dramatic consequences for all computer filesystems and ZFS is no exception. No technique used in ZFS (or any other filesystem) is capable of protecting against bit flips. Consequently, ECC Memory is highly recommended." It's not as dramatic as portrayed (probably), but again, there're 2k$ worth of drives, over 80TB. ECC RAM is negligible in terms of cost. --- How important is transcoding? Why not run the original video stream over the network? Too heavy? Device compat? 4k → 1080p?

[@justacomment1657]

@@VADemon Hey there, well zfs is going to double check every operation with checksums {if you do not disable them) . It can even use checksums to validate in memory data before writing. If you have a raid volume you will have redundant data across the pool, theses blocks house identical data and if you write or test these blocks against each other it is very unlikely that all generated checksums of all those blocks are stored in the same memory address... Alsow Sha256 is not the only algorithm you can choose for integrity checks. There are plenty to choose from. There is a nice read online hit Google and see WILL ZFS AND NON-ECC RAM KILL YOUR DATA? Published by Jim Salter

[@Aturnadagar]

Thanks for doing this build, you help me add plex to my system. I use btrfs and kvms, but beside that, I could follow the video and did similar thing, by creating a fedora kvm under my Arch host.

[@AmmarossPlays]

Glad it helped! BTRFS was a close runner-up to ZFS for this.

[@OfficialNakatsuMegami]

I love this video! Thank you for making it.

[@elstondias9172]

Great educational video 👍 I hope you make more such videos in the future featuring linux distros

[@VADemon]

Kirk you did a better job pitching this video in the podcast than you did at the beginning of the video: - What you are building (+-) - Why you are building (+-) - What will be on it and what purpose it will serve (-) - What you will show in the video (-) So beyond a few dedicated subscribers, this video will have a very hard time keeping an opportunistic viewer (from search or recommendations)

[@AmmarossPlays]

Yeah, I'm still working on formulating my delivery. I had more opportunity to think about it for the podcast, as the first third was recorded prior. I did add on-screen timestamps at the 1min mark though. :)

[@danield.7359]

I found the enclosure interesting. Probably going to use it for my next server build.

[@AmmarossPlays]

It's a great enclosure for the price.

[@peterarnfield7300]

Depends on your exact use case, but if this looks pricey, a machine to act as a Plex and gaming server (e.g. Minecraft) for a single household can be built for under £150 - I use an ex office PC with a quad core, 8gb ram, SSD and a couple of data drives running windows. Idles under 40w, we turn on when needed, and a task auto shuts the machine off at 11.30pm if we forget - been meeting our needs fine for a couple of years.

[@AmmarossPlays]

Yeah, there's ways to do this a whole lot cheaper if you're only running two drives with Windows software RAID1 (or at the very least, have backups of the non-RAID drives).

[@peterarnfield7300]

@@AmmarossPlays Yes, maybe I'm a low data user but my 600 movies, 500 TV shows and 10 years of family pictures only takes around 1TB storage, and as its not really volatile, I just use an external backup drive.

[@807800]

Any reason why Plex instead of Jellyfin?

[@timgels2918]

I can highly recommend Jellyfin for people wanting a FOSS alternative to Plex.

[@yadsarwat6325]

This is how you build a server!. Lovely job

[@Ninque]

Have this setup on unraid, but this was very interesting none the less! Great job guys!

[@VADemon]

Setting 3600 MHz XMP is not the greatest idea for a 24/7 media-file server due to increased power consumption (Infinity Fabric). Depends on your kWh price...

[@LampJustin]

Btw you can just directly install a rpm from a link (dnf install whatever)

[@potatoesarelyfe4706]

> where is Jim? reee > has been explained that he is not doing tech videos for literal months

[@AmmarossPlays]

[@IRofledTheL0L]

Ok, this is cool. I also really like the software stuff. I have a similar setup using storage spaces on windows. Obviously I miss Jim and his analysis, but Overvolted and this is great too. Keep it coming. And if I may suggest a recurring topic: Cool software and features no one knows about. Jim did a video on Radeon Chill, which got me to getting an AMD card. And I suspect there is lots of that out there, not just from AMD. Thanks Kirk for keeping this channel alive

[@AmmarossPlays]

I should do a Storage Spaces one too...

[@bakadeshi_aunstudios]

I built a 4 terrabte storage solution using freenas a few years ago. I also have a vmware box with a plex server on it. Still barely used 20% of the storage years later. I also used rosewell server chassis for both my boxes. Good cheap cases for a simple home solution.

[@gregclare]

Much prefer unRAID for a home media server solution! unRAID has the advantage of being power efficient by not requiring all drives spinning (only the drives required are spun-up). Also, continuously upgradable with larger individual drives, but still dual drive failure secure. And you can still include Plex server onboard, via Docker. I’ve been running my now 16 drive unRAID server for many years, only adding larger drives as more space is actually needed. So unRAID also allows budget optimisation by enabling you to only buy the current optimal per TB cost drive size, to add, as and when you actually need to add capacity. So your media server can just continue to grow in capacity as needed over the years, with the smallest older drive eventually swapped out for latest larger drive, as they near end of life or more capacity is needed. Nice! :-)

[@AmmarossPlays]

We set the "autoexpand" option for just that reason: as time goes by and you upgrade the size of your drives, it will autoexpand up to the size of the smallest drive in the array. You are correct you can add additional drives as your chassis allows, which is a great feature of unRAID. Since ZFS is a software RAID, it technically can allow all the disks to spin down when idle too. At least the unRAID license is fairly cheap, so cost isn't a huge factor.

[@Tempest12322]

should have put the raid card in the first pci-e slot, ryzen apus only have x8 lanes, so your not getting full bandwidth for the raid controller

[@AmmarossPlays]

Ah, true. I'm too used to their non APU CPU ;)

[@justacomment1657]

True but. A pcie 3.0 x4 slot can support 4GByte/second (32GBit). I highly doubt that those sata discs will sustain that transfer rate. If you are lucky you might hit 200MB per drive.

[@Tempest12322]

@@justacomment1657 the LSI controller is only pci-e 2.0, so depending on how many drives connected to it, you could be limited by the pci-e bandwidth, after content loading you probably wouldn't hit it,. ive got some of the seagate nas drives and they can do 230MB/s.

[@justacomment1657]

@@Tempest12322 230 sounds nice. But is that stand alone or in raid use? Haven't figured they went this old with the lsi controller, my bad...

[@AmmarossPlays]

@@justacomment1657 The LSI raid card is only PCIe 2.0.

[@TorbjrnViemNess]

Ayy, SAS2008 gang! Also nice and chill music for the assembly part

[@CriticoolHit]

Curious as to why you chose fedora over CentOS directly. Not that one is terribly different from the other. Personal preference? Thanks.

[@WallyJ2K]

Great video!!! Thanks! I may have missed it, but pricing info would have been great as well!

[@kenmorris2858]

Great video. Did you use ECC ram? Cheers...

[@BigHeadClan]

Fixed? I hear ticking from the drive still! 90% sure one of those two drives are going to fail pre-maturely but hey regular HDDs are cheap so not a big deal.

[@AmmarossPlays]

The chirps stopped when I swapped the drives. You heard the drive spin up and do an arm motor test at the end of that clip.

[@BigHeadClan]

@@AmmarossPlays Hmmm still think I heard a slight ticking, but I can't rule out the possibility I miss-heard it given how much KZbin compresses the audio. But either way its not a big deal if a drive does fail a bit early, simply the way of things when dealing with any type of computer system. Many times more so when dealing with servers. XD

[@AmmarossPlays]

@@BigHeadClan Yep, that's why we set up a raidz2 on drives with a 5yr warranty.

[@Alex-kh8zj]

@@BigHeadClan having 1 bad drive can actually be beneficial, that one will fail early so it's fail cycle will be offset

[@ThisRandomUsername]

Others may be disappointed that Jim's got other interests, but I'm just glad you guys are putting out great content like this. I'm tempted to do a much smaller version of this as a small media box also running some home automation server. For those unhappy that Jim got burnt out from a lot of the toxicity around his comment sections, maybe you're not helping.

[@AmmarossPlays]

We have a microserver build design on the website that uses a setup very similar to this, just without all the storage build-up. Handles VMware.

[@mumin84]

You could have put the RAID card in the first slot, so you'll get the CPU lanes.

[@RuSalas]

Can you migrate from trunas to a setup like this and take advantage of hardware acceleration?

[@jacoavo7875]

This was really cool. One day when I move into my own place Ill do this !

[@abstraconcept]

22:26 How does ZFS identify that a connected drive is a replacement? Without knowing exactly how it determines this, this sounds like it could have dangerous edge-cases. If it watches the by-id path you gave it, then a replacement would never appear in the same area; if it is just looking at physical connections, what happens if two drive bays get swapped, will ZFS think this is a twin drive replacement and start resilvering the two "spares" into each other? Any edge cases where the boot drive might start looking like a "spare"?

[@AmmarossPlays]

ZFS auto-replaces the drive by knowing the sdX that went missing and detects when a new device is plugged in to the same sdX. Basically, you pull the old drive from the hotswap and plug in the new drive and ZFS knows it's to replace the old one.

[@AmmarossPlays]

To answer the "drives swapped" question, ZFS can identify the drives properly, even if you pulled all the drives out, scrambled them up, and plugged them back in (obviously don't do this while it's online!). It will detect the drives as members of a ZFS already and won't purge them. You obviously need to put in a bare, unformatted drive with no partition data on it for the autoreplace to work properly.

[@abstraconcept]

@@AmmarossPlays Is there no chance that the mainboard or LSI controller might end up causing a new sd_ entry to be made rather than a new one? What if a drive is replaced between reboots? I am not entirely clear on how the drive enumeration works on Linux, but from what I remember reading one should not trust sd_ to always remain the same between boots.

[@AmmarossPlays]

@@abstraconcept if the disk is missing after a reboot, Linux will renumber the sdX to close the gap, so you'd want to do the disk swap without rebooting the server. The sdX potentially relettering on reboot is partly why we added by-id instead.

[@tlwbob]

Stacking those hard drives started to trigger me!

[@Brodda-Syd]

Drive heads are in a safely parked position. Though I would never treat them like that.

[@kborak]

Parked drives? Really? I have dropped parked drives and they are fine. You guys act like they are built of glass.

[@kborak]

@@Brodda-Syd You act like an 80's it guy.

[@Brodda-Syd]

@@kborak In 2015 one of my engineers casually dropped a draw with 8 discs on a table instead of placing gently. Within a month 3 of those drives had failed. I speak from experience.

[@kborak]

@@Brodda-Syd stop buying wd blues

[@kojack57]

I know they are out there, but, some can teach and are easy listen to and explain what they are doing. A lot can't. You fit into the former. And so, how about a fairly deep dive into ZFS and a more in dept install than this one. Please?!

[@AmmarossPlays]

ZIL, snapshots, custom compression settings, dedup, exports! So much to cover!

[@kojack57]

@@AmmarossPlays True, But, let's assume people have watched seminars, lectures and presentations from individuals and indeed the two guys from Sun who came up with it and have a thousand yard top down view of the terminology. Say someone like me, someone who is a visual learner. For example, friends don't let friends use dedup. The one thing which is difficult to find on YT is an actual walk through of actual real world configuration. Just from you creating a zpool and the commands you used WHILE creating the pool was a Eureka moment for me as I was trying to figure out in my head how those very cool features could or would be implemented. Sorry for waffling. Assume a basic knowledge and then apply a average working mans...sorry, persons implementation of ZFS. With a good title you'd get a good many views. Actually, screw views, sorry. I didn't mean it like that. You would help so many who want to use this wonderful and really cool tech. That isn't something one will often hear said about storage or volume managers. Jesus Christ, I am going to make this already loquacious longer. I know there is serious time, effort and money involved. So perhaps leave links for people to go find the .org pages and read up. You do say a three, four or five part 20 minute series on the Hardware and config end and various scenarios. Really sorry if I sound like a dick. But I REALLY want to master this area. Since discovering ZFS a couple of weeks ago it has really lit a fire under my arse and I am 100% positive there are many more like me. P.S. Cool podcast the other night. Really enjoyed it. A nice broad range of subjects.

[@AmmarossPlays]

@@kojack57 I've been using ZFS and BtrFS for things like this for some time. I'd suggest you check out the companion article to get a little bit more exposition that I didn't want to burden the video with. How-To tutorials of the nature your describing (the "here's how it would really be set up in an enterprise environment") would be a consideration. Half the time, I'm just like you, learning enough about something to use it effectively, but having to move on to the next tech before I can get too deep. I forget that as cursory as I may get, it's still more than most, so it would still be worthwhile. Thanks for watching! We try to be dynamic on Overvolted so you don't get bored thinking you're watching the 9 o'clock news.

[@kojack57]

@@AmmarossPlays Grand. Thanks. I'll head over that way now.

[@justacomment1657]

For a deep dive to ZFS I would suggest the oracle manpages for zfs on Solaris 10 (as 11 is highly uglyfied to read) and the manpages of the freebsd (11 and onward) team. Zfs is way to complex to deep dive in a single KZbin video... Alone the pros and cons of certain raidlevels and permutations... Blocksizes etc..

[@davidbeiler6364]

so how quickly do the videos load?

[@guiorgy]

I'd like to recommend (and request) you get a capture card for the next time. Otherwise, thank you for the video!

[@AmmarossPlays]

Yeah, HDMI capture for the BIOS portion would have been nice. Thanks for the suggestion. :)

[@guiorgy]

@@AmmarossPlays not only BIOS, it should be able to capture both it and the OS. No need to deal with software, or switch between the two. For an example, check Buildzoid @ Actually Hardcore Overclocking channel. As someone who deals with BIOS and PC reboots, it comes in really handy

[@hammerheadcorvette4]

Fedora !!!

[@hpsfresh]

Soooo no redundancy for os?

[@zebrasinamerica]

Audio sounded a bit muted/muffled to me.

[@AmmarossPlays]

I used the same lapel mic through the whole video (including the voiceover portion) for consistency.

[@ari-mcbrown]

Could you elaborate why you chose to do plex via http? (in stead of https) Cheers, McBrown

[@AmmarossPlays]

For initial setup, the service isn't running on HTTPS (port 443), only on HTTP (port 80). I tested this in preparation for this video. Only port 80 worked for claiming the server. Once claimed, everything works fine on HTTPS though.

[@gianlucapx]

Hey guys, I know Jim has a new YT channel. Can someone please post a link to it? Thanks.

[@adoredtv]

kzbin.info

[@homemedia4325]

Please tell me the stock cooler had thermal paste pre-applied... haha... you should have shown it!!! or done a Verge and applied more!!! (with tweezers and a Livestrong antistatic bracelet) :P

[@AmmarossPlays]

Yeah, I ended up just using the pre-applied paste as we aren't concerned about edging thermals. ;) I even had some MX-4 in the video when tossing parts around if you noticed...

[@homemedia4325]

@@AmmarossPlays I loved watching the build - ty so much for the inspirational content :)

[@erikreider]

More Linux content is always better :)

[@MirceaPrunaru]

Damn, never figured out you people will do something like this =))

[@Muhsigbokz]

I do as the algorithm guides.

[@JayceVR]

tshirt link where?

[@adoredtv]

www.teeturtle.com/products/i-woke-up-like-this?variant=24993939977

[@JayceVR]

@@adoredtv thank you

[@agenericaccount3935]

I need the shirt

[@adoredtv]

The shirt is life. :)

[@Jaythre]

19:13 since Windows 1809 (I think) ssh comes with windows natively now. You can just ssh to a machine from cmd or powershell

[@AFPMPU]

Who are you and what did you do to Jim? ☹

[@DJ_Dopamine]

Jim has a new channel now... Free Thinker, go and check it out :)

[@nathanscarlett4772]

@@DJ_Dopamine thank you!

[@uzefulvideos3440]

I think "resilvering" is only used for RAID1 or RAID10 arrays. Because resilvering is something you do with actual mirrors.

[@justacomment1657]

Actually no. In zfs you resilver the pool, does not matter if it is a raidz1 or raidz1.stripe, raidz2 or whatever someone's brain decided would be the correct layout. Resilvering will check all written data and redundant data of the pool. Pretty handy if you had a sudden powerloss. Alsow helps to identify faulty drive controllers if you run it at regular intervals. Edit: a planned test of the pool is called a scrub, Resilvering is the automatic thing...

[@AmmarossPlays]

@@justacomment1657 Yep. Just fire off a scrub on a regular basis and if it finds errors in the pool, it repairs the data. If you're replacing a whole disk, it's a "resilver."

[@justacomment1657]

@@AmmarossPlays you are absolutely correct. I misspoke.

[@justacomment1657]

Zfs check Linux... Ahhh Fedora... Ahhhhh 😰 On a server? Poor ZFS :( I hope it will last you guys, have not had great experience with ZFS on Linux, neither Ubuntu nor Fedora or rehl. FreeNAS with the plex plug in would be a good alternative.... So you can run zfs native :) ZFS compression omitting certain file types is new to me...are you sure on that? ZFS compression does alsow work if zfs is used as blocklevel device.. so omitting files by type seems strange. But maybe you are correct I haven't checked on the media server in our company... Still nice to see ZFS getting a bit more public love lately :) I really hope the freebsd guys manage proper iser integration (as host someday)... Iscsi without rdma is so slow..

[@AmmarossPlays]

I've been using ZOL (now OpenZFS) for more than 10yrs. As for compression, I think you're correct and I was confusing it with BtrFS which does have that ability.

[@justacomment1657]

@@AmmarossPlays OK, we haven't had serious problems but annoying ones. For instance: Rhel was unable to mount the zfs pool after a reboot automaticly. Problem was the system was configured to share the pool via iser. Iser could not find the mount point so the service did not come up with it... Hilarious crap. That bug is known for multiple years and the bug report has bee closed by RedHat as there is a workaround where you have to manually create some sort of cache file to tell the system that there have been a pool and it should mount it... Ubuntu was running okish with openzfs but kept sharing data via iscsi instead of iser (had to manually restart the iscsi server to load the correct configuration..) that was the only distro we had running semi productive.. Seriously..... This is just bad. Oh and in one instance - we were benchmarking different os/pool layouts the rehl distro crashed randomly. And somehow managed to corrupt the whole pool... It was a complete write off... Freebsd in the other hand. You boot it, configure it. It works.

[@AmmarossPlays]

@@justacomment1657 Guess I've been lucky with Fedora then. I use that or FreeBSD exclusively when dealing with ZFS, depending on need.

[@cdurkinz]

I know ya favor AMD generally over here, but this is one of the few dumb use cases for it. Or one of the few good use cases for an Intel CPU with their iGPU depending on how you want to look at it. Really weird to go AMD for Plex.

[@SaberOW]

Agreed, I'd definitely take an Intel CPU with quick sync for a media server. Those things are beasts for transcoding.

[@Benjamin-pg4ro]

You should have updated the BIOS first. See kzbin.info/www/bejne/Z6uWaKGCosyFfrs V1.70 was 2 month available when this was published. Because of this the XMP profile most likely doesn't work.

[@neonzz94]

You realise that Zen 2 isn't supposed to have a triangle it's supposed to have a dot the triangle indicates a fake cpu

[@AmmarossPlays]

CPUs have the dots, but the 3000-series APUs still used the triangle.

[@audy6947]

This is not why I subbed here. We need Jim's analysis. Unsubbed.

[@larryfleming7295]

What a Dick , UnSubbed? Because of this video? You obviously didn’t appreciate Jim in the first place .

[@ARCANEmateCLAN]

@@larryfleming7295 Jim retired, but he has a new channel for science related subjects.

[@itsevilbert]

The rule of thumb for ZFS is 1 GB of RAM per TB of actual disk. So does this server have 12x8TB, so about 96GiB of RAM ? EDIT: 14:51 shows 16 GiB of RAM, I guess with an absolute nightmare situation you could transfer the disks to a different system which has the needed 96GiB of RAM and rebuild there. Or possibly the future price of RAM (when a major failure does occur) will have dropped that you could, if the motherboard and firmware supports it, put 24 GiB modules into each of the four memory slots.

[@AmmarossPlays]

Your "rule of thumb" is for enabling deduplication, not for our use-case.

[@itsevilbert]

@@AmmarossPlays For deduplication the rule of thumb is "5GB of RAM for every 1TB of disk". The real problem with low RAM is when the pool is nearly full, yes it can re-silver with much less RAM, but it will require much more disk access and the one time you really do want to minimise disk access is during a major re-silver. Try filling your disks 90% full (which would never happen in real life, because finance would never delay a Purchase Order for two quarters), remove one disk to simulate a disk failure, and see how long it takes to re-silver.

[@JohnnyWednesday]

Why did you think YOUR audience wants to watch build videos? most of us are programmers and engineers - we've built EVERY computer we've owned since the Amiga

[@AmmarossPlays]

That's why I kept the build portion to the first third of the video. We stepped into beginner ZFS and Plex, which was more of the intent.

[@JohnnyWednesday]

@@AmmarossPlays - Sorry I'd just gotten up. Yeah ok - that's worthy of the channel's ethos - you may continue ;)

[@octapc]

This is not for beginners, too many words us beginners are not familiar with.

[@AmmarossPlays]

Check the companion article to see if it explains the words a bit better. It's linked in the description.

[@tomstech4390]

4 views eh? What do the comments say....

[@Aleblanco1987]

where is jim? who are you? at least have the decency of explaining what's going on

[@AmmarossPlays]

We've explained it repeatedly in our Overvolted show. These tutorials and news bits are designed for the content at hand, not to continue repeating channel news.

[@Aleblanco1987]

@@AmmarossPlays Judging by the comments that is not nearly enough. Only a fraction of the people watch overvolted. And overvolted are longer videos with varied themes. How is one supposed to know wich one to watch?. A short concise video explaining the situation would suffice. (and or write something in the community section of the channel).

[@AmmarossPlays]

@@Aleblanco1987 Correct. Jim has had the intention to make just such a video since he started Freethinker, but he's been rehashing his script for it for some time to get it just right. We all recognize it's long overdue, so we've been putting out the info in the formats that makes sense until he can get that official video made.

[@octapc]

Sorry, I cannot continue watching this channel. Good luck.

[@ClaireReed]

Alright Guys, then goes American, laughs.

[@json868]

Unraid ftw

[@crispyjelly2485]

can you at least put this shit on a fresh channel? or archive the old videos on a separate channel? I liked listening to them to fall asleep.

[@AmmarossPlays]

Jim's videos have their own playlist. Just hit shuffle on that. :)

[@ambassadorofpain1]

P R O P R I E T A R Y G A R B A G E

[@doublecrossedswine112]

What is a Linux? LoL

[@807800]

Any reason why Plex instead of Jellyfin?