[Linuxserver/plex 'site can't be reached' error]

Where is your transcode directory set to? 

 

I set mine to a cache drive folder a couple of days back while trying to diagnose a large docker image.

 

Plex web service said the server was down and I couldn't connect locally.

 

Set transcode back to /tmp which I read was memory and it's working fine again now.

[Help me sort out my dockers!!]

Looks odd to me, if that is in the sonar log it shouldn't be seeing the external IP 192.x.x.x

 

For example my logs only show:

19-12-22 17:55:09.7|Info|OwinHostController|Listening on the following URLs:
19-12-22 17:55:09.7|Info|OwinHostController|  http://*:8989/
19-12-22 17:55:10.7|Info|NancyBootstrapper|Starting Web Server

 

 

My docker config looks comparable

 

 

 

In the docker I just have this for the port

 

 

 

[New harddrive with read errors [Solved]]

Segate does it differently to WD, they report the number of reads without an error the opposite of WD. 

 

My Unraid segates are:

                                  Current  Worst  Threshold   Raw

1  Raw read error rate     100      064     006           2301087                     ~13000 hours power on

1  Raw read error rate     081      064     006           134297756                 ~6000 hours power on

 

Both in good health, the stat gets closer to 100 as the drive ages, until you start to get errors and it drops again.

 

Two other segates in my about to decommision WHS2011 build which have worked a bit harder now have a current of 115 and 117, both with >19000 hours.

 

[so Im looking to build a bigger low power nas]

5 hours ago, suender said:

I have exchanged my Ryzen 1700 (over 70 Watt, no bios modifications) into this system since yesterday: J5005. (4 Cores)

 

I have with 2x10TB HDD and 1X500GB SSD only 24 Watt (IDLE). Maximal Performance I get 32 Watt.

 

Those processors offer amazing performance per watt, actual perfomance almost on a par with the old 65W TDP I5 2500s I have in a H61 mITX board but at a fraction of the power usage under load.  The only issue for me is the platform has such limited upgrade options so you have to be confident you won't outgrow the system.

 

If I was spending my own cash on a new budget build, I'd mostly likely go for  Pentium G5400 in a uATX B360 motherboard. 6 sata, 4 standard dimms, PCI-E x16, x1, x1. CPU is faster in single thread and mulithread while peaking at 25W underload. You'll have to add in some platform and system losses which will bump it up a bit, but still I'd take the flexibility over a few extra watts.

 

 https://www.anandtech.com/show/13660/amd-athlon-200ge-vs-intel-pentium-gold-g5400-review/20

 

Just goes to show, how low we can go.

[Old Unraid server locking up. Repair or build new?]

Hi,

 

If you have multiple Dimms then it's unlikely more than one has failed. Test them independantly, you may be able to get the system running which will make transfering files easier.

 

Given the vintage of the hardware, I'd recommend you update to something a little more modern so you can run the latest unraid. This doesn't need to be expensive. A dual core pentium, basic motherboard and 4GB ram will be a significant upgrade and even when bought new should cost less than $200. I'd also get a new PSU if it's of similar vintage.

 

The onboard SATA and NIC should do just fine in any recent mainboard.

 

Alternative is look out for a cheal HP N36L, N40L or N54L second hand which you could pick up for less than $100

 

 

[Disabled disk (Red X)]

Looks to me like the disk is toast.

 

187 Reported_Uncorrect       -O--CK   001   001   000    -    65535

197 Current_Pending_Sector  -O--C-   001   001   000    -    17480
198 Offline_Uncorrectable       ----C-   001   001   000    -    17480

 

Compared to one of my seagates - ST4000LM024

 

187 Reported uncorrect                  100      100   000      -  0

197 Current pending sector            100      100   000      -  0

198 Offline uncorrectable                100      100   000      -  0

 

That model drive is know to have issues https://en.wikipedia.org/wiki/ST3000DM001

[so Im looking to build a bigger low power nas]

As S80 says, the processor uses a lot less power at idle, power consumption scales with voltage and frequency. As a rule, you typically need more voltage to run at a higher frequency.

 

At full load a 65W CPU may be running at 4Ghz and 1.3V while at idle it can drop to 800Mhz and 0.8V so may only need 5-10W.

 

Sounds like you have a lot of hardware choices. If power consumption is a signifcant factor, I'd stick to something relatively simple, a big dual core server with onboard raid controllers and noisy fans could easily idle at 100W+ and be overkill for personal use. (though whoever said they have too much capacity)

 

Unraid is system agnostic so will boot in almost anything. You can move the USB stick and drives to a totally different system and it should still boot. Take a USB stick with trial Unraid installed and a killawatt meter to work and test a few systems you like the look of. Drive power consuption relates to the number and types of drive so taking a base line of a driveless system is a good comparison point.

 

One other thing to note, Hard drive power consumtion is mostly linked to the number of drives so filling your server full of repurposed low capacity drives would need a lot more power and a bigger PSU to deal with start up surge (2A per drive) than a few high capacity drives.

 

 

[so Im looking to build a bigger low power nas]

Hi,

 

Anything from Intel second gen has decent performace and low idle power consumption so 4th and 5th gen gives you plenty of options.

 

A good few years ago I imported a reasonably priced 25W 1.6Ghz dual core Athlon II 250u and was happy with the idle power consumption. Later this was replaced with a 95W 3.0Ghz quad core Athlon 640 X4 in exactly the same system and the idle power consumption actually dropped a few watts!

 

Low wattage CPU's are mostly just capped in the clocks they can boost to so stay at full boost for much longer to get a task done before dropping back to idle, overall power consumption is not that different. Unless you have specific ongoing workloads, your unraid nas will be near idle most of the time so idle consuption is the focus. Even if you are transcoding, you can usually pass that off to the iGPU so the CPU doesn't have a lot to do.

 

My current Dell T20 E3-1225V3 is a 4th Gen haswell, with 8 x 3.5" drives spun down, 4 x SSD active, two 8 port LSI raid cards and an external SAS 4 bay enclosure with independant power supply on the same UPS I'm at around 65W idle. WIth everything spinning closer to 90W. Before I added all the raid cards, extra drives etc. Annual electricity cost £90 ~ $110. For info, the Dell idled around 40W with a couple of drives.

 

If you can find something with iGPU, I3 2C4T or I5 4C4T and the most SATA ports you won't go far wrong to get set up. The issue you may have is proprietry PSU's, motherboards, cases etc which could limit expansion but if you get any server hardware in, picking a tower with a reasonable number of drive bays and a CPU comparable to desktop such as the E3 series should net you a reasonablly efficient system.

 

One thing to avoid is filling it with lots of small repurposed HDD's... this will soon add up as power consuption. Less is more when it comes to spindles, and if you can strech to a reasonable SSD for the cache drive, the array can sleep most of the time with the SSD only pulling a couple of watts.

 

 

 

 

 

[LCC and drive durability]

Hi,

 

TL:DR My drives have LCC of 1.5m and 850k vs data sheet 300k. Is this really an issue?

 

I reconfigured my unraid around 2 years ago with new and repurposed drives and then about 12 months ago added a second parity and second SSD for cache.

It's pretty much been left to its' own devices save periodic OS updates and adding more dockers. Over this holiday period I swapped both 4TB parity drives for 8TB.

 

While looking in the smart logs, I see the the 3TB Green WD30-EZRX has and LCC of almost 1.5m and a shucked 4TB from early 2019, I suspect a white labeled blue WD40-EMAZ has an LCC of almost 850k. Both of these drives are speced at max LCC of 300k and the smart data shows 001 (starts at 200)

Neither drive has ever flagged a read error or a SMART error.

 

Turns out my docker and libvirt images were on drive 1 (WD green) and the writes were happening constantly, but not constantly enough to stop the heads parking every 8 seconds. The parity drives were a 4TB Red (still low LCC)  and the shucked blue which also must be parking at 8 sec. I've now moved the files to the cache and my array now spins down fully when not in use. I updated parking to 300s on the Green drive, however none of the tools worked on the blue so I have that to figure out yet.

 

So the question is, is the high LCC count still a significant risk to the drive integrity?

 

The 4TB blue? is now unassigned. I've now run a full pre read / zero / post read, pre-clear cycle and an extended smart diagnostics,  SMART data is still good, not a single replaced sector, all paramters normal except the super high LCC. Currently this is out of the array as I'm pre-clearing another 8TB I shucked.

 

The 3TB green is still in service, with 4TB Red pre-cleared  and on standby. Again the 3TB has no smart errors/ warnings or replaced sectors despite the 1.5m LCC. Once the 8TB is cleared, I could replace it with the 4TB red which would allow me to test it out.

 

While the safest option may be to repurpose these drives for something else, is this LCC really such an issue, or just what was tested in the validation. Early drives with high LCC had known elevated failure rates, however given the Green is at 1.5m it gives me some confidence I should get similar mileage from the blue and both drives so no signs of issues at this point so it seems a shame not to use them. Even the Green isn't likely to increase the LCC much now it's fixed.

 

Anyone have experience of running drives with such high LCC ?

 

Topic posted as while I can see lots post concerned with LCC, I can't find any on whether there were real consequences or examples of people running them with high LCC.

[Last built a PC in 1995! Sanity check please.]

Hi,

 

Have you looked at the SilverStone SST-DS380 case.

It is a bit more expensive, but you seem to gave a good budget.

8 x front removable 3.5" / 2.5" and 4 x internal 2.5"

Only slightly larger than the Node 304.

You can add a controller card in your PCI slot if you need more than 6 drives

 

Once I had 2 x Sata cache , 2 x parity I soon needed more drives

 

Edit: Note it needs an SFX power supply 

[Dell PERC H200 card on PCIE 2.0 x1 slot via extender cable ?]

Hi,

 

A single SATA 3.0 comes close to saturating PCI-E x1 2.0 so this isn't likely to be a great solution.

 

You may get 2-3 spinning discs without a performance hit, but any more than that will be a bottleneck.

 

[Unraid server low power]

Hi,

 

I've only been on unraid a couple of years but have had home servers I've tried to keep efficient for the last 15 years.

 

Most Intel cpus are very efficient at idle so little to choose between them unless going for atom class parts, just spec what you think will last. 6C 6T should be good for many years (though 2C 4T would probably use a little less power at idle) and if you do want to use transcoding the IGPU takes the load once configured that way.

 

Motherboards are a personal choice, my old WHS ran for 10 years on a consumer grade MSI board. The board is still fine now, repurosed for flashing cards, drives etc on the bench.  Server boards may prove more reliable though without any remote access features you're paying a premium for a base spec. I'd either pay less for a consumer board, or more for a board with remote features.

 

PSU is where I'd differ. I far prefer a conventional PSU for multiple spinning drives and you never know when you may want to expand to add extra contreller cards etc. 3.5" disks can pull up to 2A during spin up, when your array of 4 drives starts to spin up, thats a lot of surge for a 12V inline adapter to handle along with CPU demand with little space for smothing capacitors etc. This can be for a parity check or any other purpose.

 

You'll get a good brand 300-500W gold supply in the same budget. 300W is likely to be more efficient at low wattage.

When I did some testing a few years back with an I5 2400s in a mini ITX board and using a good 12V adapter, the Pico 120 idled around 22W while a Huntkey Gold 300W PSU idled around 4W higher. Note the Huntkey wasn't what I would call a quality brand but it was never under any stress so is still fine now. 

 

In my view, the Pico's and similar are great for many small build purposes, but when it comes to mutiple spinning drives, I'd much rather have a purpose build PSU with some reserve especially when the power consuption difference can be minimal.