TR Forums

I have an i5-4430 in my Plex server and it just manages to decode H.265 1080p content. It just gets an occasional rest from being maxed out every few seconds but it is consistent.

techguy wrote:

As for actual reliability, I've had those 4TB drives in service for almost 2 years now with no failures of any kind. I'm beyond impressed with Toshiba. At this point I only recommend HGST and Toshiba when someone asks me for a storage recommendation. Seagate can pound sand for all I care and WD is usually over-priced.

my 2 cents

Anecdotes are fun. Seagate almost always ends up in nearly the same AFR categories as Toshiba and HGST/WD. There was a bad run here and there, but that happens to every company occasionally.

Waco wrote:

techguy wrote:

As for actual reliability, I've had those 4TB drives in service for almost 2 years now with no failures of any kind. I'm beyond impressed with Toshiba. At this point I only recommend HGST and Toshiba when someone asks me for a storage recommendation. Seagate can pound sand for all I care and WD is usually over-priced.

my 2 cents

Anecdotes are fun. Seagate almost always ends up in nearly the same AFR categories as Toshiba and HGST/WD. There was a bad run here and there, but that happens to every company occasionally.

Yeah, about that...

https://www.backblaze.com/blog/hard-dri ... s-q3-2017/

techguy wrote:

Yeah, about that...

https://www.backblaze.com/blog/hard-dri ... s-q3-2017/

I'm not sure what you thought you were posting, but that backs up my assertion.

Waco wrote:

techguy wrote:

Yeah, about that...

https://www.backblaze.com/blog/hard-dri ... s-q3-2017/

I'm not sure what you thought you were posting, but that backs up my assertion.

2 Seagate drives on that list with failure rates > 10%. Zero from any other vendor.

Waco wrote:

End User wrote:

Ah. Those X300's aren't NAS drives so they are cheap.

NAS drives are overrated, IMO. Unless you're sticking them in something you intend to call support for (where the qualified drive list may get you into trouble if you deviate) I wouldn't spend the extra cash.

It is tempting to go cheap.

I'm at a crossroads now. My current NAS setup is 5 years old and I am running out of space. It is going to be pricey if I stick with NAS rated replacement drives (8 or 10 TB). I must ponder further.

techguy wrote:

End User wrote:

techguy wrote:

Storage is cheap. I have 56TB in my Plex server.

What is your setup?

10x 4TB Toshiba X300 drives + 1x 8TB Toshiba N300 + 1x 8TB HGST Deskstar NAS drive = 56TB (not using RAID, separate external 8TB disks for backup)
i9 7900x @ stock under custom water (3x120 + 1x140mm rads)
Asrock X299 Professional Gaming i9 (integrated 10G NIC)
32GB DDR4 3000 (4x8)
Corsair 900D
PC Power & Cooling Silencer III 1200W

I have room for one more 3-drive caddy if I downgrade my PSU to a smaller unit which brings the chassis up to 15 drive capacity. I'll probably do this in the next 3-6 months.

OS?

techguy wrote:

Waco wrote:

techguy wrote:

Yeah, about that...

https://www.backblaze.com/blog/hard-dri ... s-q3-2017/

I'm not sure what you thought you were posting, but that backs up my assertion.

2 Seagate drives on that list with failure rates > 10%. Zero from any other vendor.

Yeah, with super low quantities and low confidence.

You'll note that they're generally all 3% or less with few exceptions.

Honestly your best bet is to use a client that can decode HEVC, they are under $100 not to mention the power savings. I use a WebOS app that can Direct Play HEVC, otherwise I'd look into Roku, Odroid, or Rock64.

A good read from the Plex forums: https://forums.plex.tv/discussion/28289 ... ranscoding

I have Plex running in FreeNAS in a ESXi box. It's not that bad. The CPU's are the low-power Xeon 6-cores, 12-Threads from years ago. I have a 8 "cores" assigned and unlimited resources (10% system wide reserved for the router, though). I can transcode 2-3 streams at once, enough to transcode while recording an entire HD Homerun Prime. With a bit more core assigned I think it could handle H.265. About the power, yes, it's a lot.

If I could do it again I'd go with a Ryzen and a GPU for transcoding. Neither were an option a year and a half ago when I started down this path.

End User wrote:

techguy wrote:

End User wrote:

What is your setup?

10x 4TB Toshiba X300 drives + 1x 8TB Toshiba N300 + 1x 8TB HGST Deskstar NAS drive = 56TB (not using RAID, separate external 8TB disks for backup)
i9 7900x @ stock under custom water (3x120 + 1x140mm rads)
Asrock X299 Professional Gaming i9 (integrated 10G NIC)
32GB DDR4 3000 (4x8)
Corsair 900D
PC Power & Cooling Silencer III 1200W

I have room for one more 3-drive caddy if I downgrade my PSU to a smaller unit which brings the chassis up to 15 drive capacity. I'll probably do this in the next 3-6 months.

OS?

Win 10 Pro. It's just a Plex server/media workstation. I don't need a server O.S. for that and with the editing work I do I might as well just stick to Windows.

Waco wrote:

techguy wrote:

Waco wrote:

I'm not sure what you thought you were posting, but that backs up my assertion.

2 Seagate drives on that list with failure rates > 10%. Zero from any other vendor.

Yeah, with super low quantities and low confidence.

You'll note that they're generally all 3% or less with few exceptions.

Toshiba and HGST drives are in the < 1% failure rate range, so even 3% looks not so great.

Buy whatever drives you like, I'll stick to Toshiba and HGST.

techguy wrote:

Toshiba and HGST drives are in the < 1% failure rate range, so even 3% looks not so great.

Buy whatever drives you like, I'll stick to Toshiba and HGST.

I'm not saying you shouldn't, I'm saying if you even remotely protect your data in a reasonable way the AFR doesn't matter at all when they're all so low.

You'll note that the vast majority of drives that Backblaze owns, even after the 3 TB drive fiasco, are Seagate.

Waco wrote:

techguy wrote:

Toshiba and HGST drives are in the < 1% failure rate range, so even 3% looks not so great.

Buy whatever drives you like, I'll stick to Toshiba and HGST.

I'm not saying you shouldn't, I'm saying if you even remotely protect your data in a reasonable way the AFR doesn't matter at all when they're all so low.

You'll note that the vast majority of drives that Backblaze owns, even after the 3 TB drive fiasco, are Seagate.

I backup all my media to external 8TB drives.

I imagine Seagate has fairly aggressive pricing in light of past events. Many CIOs care about price first and all else second, in my experience. Not surprised to see Seagate's marketshare hasn't dried up despite their reliability issues over the years. It's not just 3TB drives they had issues with, that's just the most recent example. 7200.9 and 7200.11 drives suffered extremely high failure rates as well.

techguy wrote:

Many CIOs care about price first and all else second, in my experience.

Many CIOs are pretty bad at making decisions about storage technology.

I've got ten Hitachi desktop drives in my Plex box, ten 2TB of various sizes in my FreeNAS, and nice certified 15K 450s in my ESXi machine. All are about the same reliability. Only the warranty speed matters. That's why my Hitachi setup has a cold spare. On disk fail, replace with spare, begin resilver, begin RMA. I'll have a week or two of no-spare, less than a day of single parity (RAIDZ2 ~= RAID6 ~= RAID5 with a second parity setup), and nearly zero downtime.

11 desktop disks plus warranty extensions cost me about 1800$.
10 WD Red disks, same size, was 2650$. No brainer call.

You don't need Enterprise support for your home lab. You don't need five nines. You need Good Enough, and for almost all uses, that means desktop drives in a big raid array is a non-issue.

As for OP; don't play with h265 yet. Add more storage and keep the complexity (and brittleness) of the system low. All my UHD movies are in h265, since the source is h265, but otherwise all my collection is h264, and for good reason. Getting h264 hardware encode and decode working is easy, really easy. Getting hardware h265 support is really, really hard. You'll also want hardware encode *AND* decode for best results, and that's all but unreachable. Hint: NVENC's h265 output is really, really subpar. Watchable for sure, but won't save much space.

And the nice thing about Plex for movie rips, is if you lose a percentage of them them, big whoop. Just re-rip them.

Forge wrote:

I've got ten Hitachi desktop drives in my Plex box, ten 2TB of various sizes in my FreeNAS, and nice certified 15K 450s in my ESXi machine. All are about the same reliability. Only the warranty speed matters. That's why my Hitachi setup has a cold spare. On disk fail, replace with spare, begin resilver, begin RMA. I'll have a week or two of no-spare, less than a day of single parity (RAIDZ2 ~= RAID6 ~= RAID5 with a second parity setup), and nearly zero downtime.

11 desktop disks plus warranty extensions cost me about 1800$.
10 WD Red disks, same size, was 2650$. No brainer call.

You don't need Enterprise support for your home lab. You don't need five nines. You need Good Enough, and for almost all uses, that means desktop drives in a big raid array is a non-issue.

I'll raise you another - of the 16 2 TB drives in my server, all of them are refurbished and none of them cost more than $50. Two spares sit on the shelf in case any failures arise.

After 3+ years of them running, I've had one failure that cost me another $50 for another cold spare to sit on the shelf (after fully testing to ensure it wasn't going to die the instant I hit it with a resilver).

Refurb drives, in my experience, are just drives someone lost a partition on and RMA'd. An industry exec (from an unnamed disk vendor) told me that 98% of all disk RMAs have zero actual disk defects...

Waco wrote:

Forge wrote:

I've got ten Hitachi desktop drives in my Plex box, ten 2TB of various sizes in my FreeNAS, and nice certified 15K 450s in my ESXi machine. All are about the same reliability. Only the warranty speed matters. That's why my Hitachi setup has a cold spare. On disk fail, replace with spare, begin resilver, begin RMA. I'll have a week or two of no-spare, less than a day of single parity (RAIDZ2 ~= RAID6 ~= RAID5 with a second parity setup), and nearly zero downtime.

11 desktop disks plus warranty extensions cost me about 1800$.
10 WD Red disks, same size, was 2650$. No brainer call.

You don't need Enterprise support for your home lab. You don't need five nines. You need Good Enough, and for almost all uses, that means desktop drives in a big raid array is a non-issue.

I'll raise you another - of the 16 2 TB drives in my server, all of them are refurbished and none of them cost more than $50. Two spares sit on the shelf in case any failures arise.

After 3+ years of them running, I've had one failure that cost me another $50 for another cold spare to sit on the shelf (after fully testing to ensure it wasn't going to die the instant I hit it with a resilver).

Refurb drives, in my experience, are just drives someone lost a partition on and RMA'd. An industry exec (from an unnamed disk vendor) told me that 98% of all disk RMAs have zero actual disk defects...

The big box, with the 4TB disks, it was purchased for the purpose with a tax return (wife approved!). All of it was new. The FreeNAS box was assembled piecemeal, I don't think more than 2 of the 2TB disks match. Many of them were 50-60$ refurbs from Microcenter. I feel much the same about refurb disks. Put it in a dock, beat it up real good for a day or do, lots of copies on/off, power cycle it a dozen times, then put it in the array and resilver or mark it cold spare and ready for use.

That's actually a very large part of why I use RAIDZ2 instead of RAIDZ1, even though most of my data is easily regenerated or redownloaded.

Nice! I keep my cold spares on the shelf, literally cold, just because I tend to prefer to swap out the drives in-place if something goes awry. I go a step further and build RAIDZ3s simply because drives are cheap. I even keep a cold backup on the shelf (on a 10 TB and 6 TB drive) just in case a few files get holes blown in them from a failure.

DragonDaddyBear wrote:

I wonder how well it would work with a newer kernel. That's a new-ish CPU for an old-ish release. Can you try swapping kernels and let us know how it works? Or maybe try Clear or something with newer drivers?

It *still* doesn't work on Ubuntu 16.04.3 HWE, which is 4.10.

It just completely locks up the computer, with nothing helpful in any of the logs.

Since this is the main server for a bunch of stuff, I'm loathe to try it again.

The only thing I can say is that whatever is happening with me doesn't seem to be very common, at least at first glance: cursory googling doesn't find anything relevant.

Sort of an odd question, but my Plex server is a headless box and all of my clients are either Roku's, a Tivo, or a couple of android phones. A gpu will still provide h.265 decoding, even if the local machine isn't displaying the video, right?

In theory as long as the system can access the GPU it should work.

Glorious wrote:

DragonDaddyBear wrote:

I wonder how well it would work with a newer kernel. That's a new-ish CPU for an old-ish release. Can you try swapping kernels and let us know how it works? Or maybe try Clear or something with newer drivers?

It *still* doesn't work on Ubuntu 16.04.3 HWE, which is 4.10.

It just completely locks up the computer, with nothing helpful in any of the logs.

Since this is the main server for a bunch of stuff, I'm loathe to try it again.

The only thing I can say is that whatever is happening with me doesn't seem to be very common, at least at first glance: cursory googling doesn't find anything relevant.

I'd open a ticket with them.

DragonDaddyBear wrote:

I'd open a ticket with them.

Plex support?



No such thing. I've been using Plex for years and as much as I love it when it works, it is literally the worst supported piece of software I have ever used. Plex has user forums but rarely does a Plex employee post and most issues go unresolved for months. New client software versions in the last couple years have broken more things than they have fixed too.

DragonDaddyBear wrote:

I'd open a ticket with them.

I would, but:

1) I don't have anything of use to tell them.
2) I''m unwilling to go through the hassle of replication because I use this server for tons of stuff.

It seems silly to waste their time if I've already precluded the first thing I assume they will ask of me. I'm fine with not having the feature anyway, I just thought it was cool and might save some power. No real loss of utility.

Also, as techguy has said, the overall support situation (not necessarily the official plex team) always seems so dismal: so many people talking about plex as a general subject seem to have no idea what they are doing. That kind of thing is very discouraging for me, because I hate having to explain, in exacting detail, a problem for which the "solution" is provided by a flood of people who are so overconfident in its efficacy that they don't even bother to check for its applicability.

TheEmrys wrote:

Sort of an odd question, but my Plex server is a headless box and all of my clients are either Roku's, a Tivo, or a couple of android phones. A gpu will still provide h.265 decoding, even if the local machine isn't displaying the video, right?

Yes and no. I don't think it's saying what you're thinking. On the plex server, the media must be decoded, and then encoded, unless the target is doing direct stream playback. A good GPU on a headless server will do hardware decoding of the on-disk format, and then hardware encoding of the target format (usually mpeg4/h264). Once that's done, the client decodes it however it likes. The server can't "hardware decode" media on the client (not what I think you meant), but it can hardware decode and then encode it, saving CPU on the server at a small penalty to image quality (NVENC, QuickSync, etc all have poor quality and excellent speed compared to a software encoder).

Hardware DEcoding, however, usually provides quality very close to or identical to software decoding, so the best image quality path is probably a hardware decoder but a software encoder. It's out of scope for today, though.