Это видео недоступно.
Сожалеем об этом.
A ZFS Masterclass with Tom Lawrence
HTML-код
- Опубликовано: 25 апр 2024
- In part 2, Tom Lawrence of @LAWRENCESYSTEMS gives us an absolute masterclass in ZFS, answers some audience questions, talks about some of his favorite open-source projects, and opines about the shift away from the cloud back to self hosting.
Did you miss Part 1?
• Getting to Know Tom La...
Interested in learning more or keeping up with Tom Lawrence?
lawrencesystem...
/ @lawrencesystems
/ tomlawrencetech
/ lawrencesystems
Show Topics:
0:30 The Billion Dollar Filesystem
3:39 ZFS Ram requirements
5:15 ZFS Data Integrity and checksums explained
8:14 ZFS Caching with ARC
18:26 Should your media shares be in a zpool on ZFS?
20:56 ZFS vdev setups and considerations
24:14 RAIDz1, z2 and z3
28:45 Expanding ZFS Storage and potential data loss scenarios
30:46 Rsync vs ZFS replication
33:24 What is the maximum recommended drives to put in a vdev?
40:12 Exploring new technologies in ZFS for business and homelabbers
50:31 VMs on ZFS: Zvols and vdisks
52:42 User questions
59:06 Best practices for setting up a NAS with ZFS
1:01:15 Favorite Open Source projects
1:09:03 The shift from cloud services to self hosting
Join this channel to get access to additional Unraid Community perks:
/ @uncastshow
![Chuckyy - Devil Hours [Official Music Video]](http://i.ytimg.com/vi/f3wZGj0VnZU/mqdefault.jpg)
Man, two of my all time favorite creators. This was a great show thanks for putting it together!
Another great Uncast session - thanks to Tom & Ed for taking the time to film, edit and upload. 👍
Good videos Ed and Tom Thanks for always sharing with us!💖👍😎JP
Have been a huge fan of ZFS since its origins on Sun Solaris. But I think you may be overstating the checksum protection as a reason not to care about ECC. If a bit gets flipped before the checksum is calculated, the checksum won't help you. It's probably a very minor issue for almost all users, but thought it worth mentioning.
I apriciate all the content both of you produce, I’ve learned so much from you guys
Excellent video, guys. Learned a lot 👍
Great show, thanks for that. Did anyone else notice Tom's subtle audio stuttering/popping?
Wow, what a nice Guy with so much knowledge! Go on, and have maybe more Free time you earned it!
Thanks for everything!
oh man I didnt know you guys released this, two of my fav creators :D
Loved the videos and podcasts!
Love the T-Shirt SIO
super nice!
I wonder what he thinks about OneFS
extremely informative talk! Thanks for this!
My body, and server, are so damn ready for version 7 :O.
Can you restrict a pool from being loaded into the cache and just have it read from storage every time? Media files for example.
For this, the best way would be to tune the pool(s) to cache very little but have several datasets within the pool, which have different caching parameters set for each of those datasets, each of which can be mounted to a different point in the same top-level filesystem (yes, all block devices are able to contain others, and a block device can be a drive, a pool, a dataset, or even a file), so the cached files for (say) VMs, video files, audio files, photos, textual files, etc. are all different, each mounted within the same filesystem tree. They can each use a different blocksize as well, for efficiency, and each dataset can utilise a different format - which as it is just a block device can be anything supported by the operating system that will use it.
I use rclone set to use 32 worker processes for senarios with lots of files.
Can someone link the video around the dual-backplane ZFS system (if I understood that right)? I'd like to hear more. I'd try to search Tom's back catalogue but I'm .... what's the word .... lazy.
I remember watching that video a while ago, but I'm obviously not going to spend the time looking for it if you can't be bothered to put in the effort yourself.
Ed his voice reminds me of Spaceinvader1. I don't know if it is the same person
Same guy! 😜
Leading up to 24:20... I couldn't tell if he doesn't know what he's talking about... or if he is just saying "mirrors" and means "stripes", or "Striped mirrors" (RAID 10). Because, adding two drives at a time, and of those two, one mirrors the other, well... either can die. For any pair. You're only in trouble if you loose two in the same pair. Thats how RAID works anyways, maybe there is ZFS jargon i'm not aware of...
So, @38:11 you had my full attention since you were describing my exact environment. Unfortunately, you never gave WHAT was the better solution for said scenario. Just using Unraid for PleX and a lot of 4K movies. Its been about 3yrs since I looked at changing over from a Windows setup with 24 drives. Now that I lost a drive and all its data...... I had planned on going to the 32 (all 8TB but a mixture of WD and Seagate) with 2 parity and using NVMe quad card with 4x 1TB drives for cache. Thanks for everyone's time and great video, again.
But you didn’t actually lose any data because you had a backup, right?
Right?….
RAID is not a backup.
ZFS is not a backup.
If you don’t want to lose data, you need to have a backup.
For media server purposes, I recommend using XFS with a standard Unraid array. XFS tends to perform better with full drives and allows for easy expansion should you need additional storage. Using two parity drives will provide solid redundancy. The advanced features of ZFS might be overkill for a basic media server-however, there is always a 'but'. If your Plex server services numerous users who might watch the same show or movie simultaneously, ZFS's drive striping and ARC cache can significantly boost performance. But if you're using a large setup, such as 32 8TB drives, you'd need to create a multi-VDEV pool comprising either three RAIDZ1 or three RAIDZ2 configurations. So, for those running a media server with few concurrent users, sticking with an Unraid array and XFS is best IMO. For the cache pool, however, I would always opt for ZFS.
@@williamp6800 Was on a Windows build and no,,,, No I didnt have any back up. I have learned my lesson and in process of building an Unraid machine.
Let’s use AI to build a fs
At least we'd get decent documentation, clear usage licensing, an honest declaration of it's intended usage along with it's strengths and weaknesses backed by sane benchmarking data.
@@AshtonClemens yea usually devs wanna do it but then you have a few special one s