BEWARE! If you're using a MicroSD, switch to a HDD NOW!

Valentino_Stillhardt · September 18, 2020, 4:39pm

I set the NUC cooler to 100%, it’s now louder, but cooler inside.

I will let it run and see if it unmounts itself again.

Currently under investigation! Please wait a few days.

I removed “SSD” from the title, since it’s not totally dead and I still have one variable to test: Heat.

EDIT: Even with the cooler at 100% blowing warm air out it seems hot:

smartctl -x /dev/sda
...
Current Temperature:                    61 Celsius

sensors
coretemp-isa-0000
Adapter: ISA adapter
Package id 0:  +54.0°C  (high = +87.0°C, crit = +105.0°C)
Core 0:        +54.0°C  (high = +87.0°C, crit = +105.0°C)
Core 1:        +51.0°C  (high = +87.0°C, crit = +105.0°C)

Bartem · September 18, 2020, 4:53pm

So, speaking of this… I currently just do a local backup and a network backup to another system in my house… I also have a private git-hub which I usually forget to push updates to. Does anyone know if there is a way to have the changes automatically pushed to GitHub with different branches so I have some history if I wanted to look back at if I made changes I am not thrilled with later?

tom_l · September 18, 2020, 4:58pm

I only know of addons that do automated backups to Google Drive, Dropbox or a Samba Share (NAS).

To a NAS: Samba Backup: Create and store snapshots on a Samba share

To Goggle Drive: Hass.io Add-on: Hass.io Google Drive Backup

To DropBox: https://github.com/danielwelch/hassio-dropbox-sync

code-in-progress · September 18, 2020, 5:12pm

Yeah, 61C is pretty hot and is likely a contributing factor. None of my servers, even a couple sitting in a wooden cabinet, rarely reach above 53C. With that said, most consumer SSDs are rated between 0C and 70C with the latter being the absolute hottest a consumer SSD can stand before starting to fail. There are plenty of reports of SSDs failing even at 62C-65C for extended periods.

If you can, I’d recommend placing even a small fan blowing across the top surface of the NUC and see if you can get that temp down further as eventually the NUC will start to roast itself.

Burningstone · September 18, 2020, 5:31pm

I think this tool should work to automatically push on any change.

Why do you need to push it to different branches? You already have history even when you push to one branch only.

Bartem · September 18, 2020, 6:27pm

Not an often git user so sorry if I don’t understand exactly … so the compare changes will be there? But just for the one previous push correct?

code-in-progress · September 18, 2020, 6:37pm

Nope, you can diff all the way back to the initial checkin of the file.

Bartem · September 18, 2020, 6:54pm

Awesome thank you, I didn’t realize it kept all of them

Valentino_Stillhardt · September 20, 2020, 1:04am

UPDATE:

Now running almost 2 days with the NUC fan blowing at 100% and SSD hasn’t unmounted yet. Also, I read some informations below and if I’m getting that right, the SSD has at least once in it’s lifetime reached 109C.
I think it might be clear why it has begun to fail. However, I am still not sure why it has never failed before. We had a couple of hot days before and it never showed that problem, just now.
Surprise: It’s still alive.

smartctl -x /dev/sda
...
Current Temperature:                    51 Celsius
Power Cycle Min/Max Temperature:     34/63 Celsius (2 days on)
Lifetime    Min/Max Temperature:     20/109 Celsius (WTF?)

sensors
coretemp-isa-0000
Adapter: ISA adapter
Package id 0:  +48.0°C  (high = +87.0°C, crit = +105.0°C)
Core 0:        +48.0°C  (high = +87.0°C, crit = +105.0°C)
Core 1:        +41.0°C  (high = +87.0°C, crit = +105.0°C)

Conclusion: Intel NUCs haven’t sufficient airflow, be sure to set the fan at 100% or near that if you plan to use it 24/7.
I bought a new mSATA 120GB from Kingston to see if the newer SSD is more efficient, and thus, cooler. The current SSD gets pretty hot. For the curious, it’s currently a “Intel 525 Series SSD”.

Long SMART test:

Num  Test_Description    Status                  Remaining  LifeTime(hours)  LBA_of_first_error
# 1  Extended offline    Completed without error       00%     47383         -

Good idea. I am thinking of printing a 3D case to screw on which has a fan mount. I would replace the top with the 3D one.

nickrout · September 20, 2020, 2:14am

If it has started to fail, I would replace it. It can’t get better, only worse.

KennethLavrsen · September 20, 2020, 10:46am

I am a hardware engineer with 32 years behind me + hobby since I was 10 years old

Hardware is not software. Hardware is made from materials. Materials change over time. Two pieces of hardware are not identical. One can fail after a month. One can fail after 10 years even if the official mean time between failure is 3 years.

If an electronic component has failed at warm temperature replace it! Temperature accelerates transformation of materials. When you do accelerated life tests the two main components in the test is hot soak while running and extreme temperature changes. A component that has run too warm has been aged in an accelerated way. Replace it while you can still get data off it

As many say SD cards are not really designed for heavy reading and writing of small amounts of data.

Both SD cards and SSDs are made from flash memory. Flash bits can be set to 0 individually but cannot be reset back to 1 other than erasing entire blocks.

So when people talk about how many petabytes you can write, then it depends on how you write. In a video camera you create gigabyte large files but you create them block by block in a stream where you add new blocks to the file. SD cards and SSD handle this very well.

But when you have software that writes few bytes to same few files over and over again the number of bytes that actually gets altered on the flash can be much larger.

And if you have an 8GB card already filled with 3-4 GB data of which a lot is dynamic data like the database then you ask for trouble.

An SSD has better wear levelling algos, it caches more data in RAM to limit writes, it is generally larger than an SD card so it has much more space to do wear levelling.

A spinning harddrive does not have a practical limit to how often you can change the magnets on the platter. But it has moving parts. The disks are spinning between 5000 and 10000 times per minute 24/7 which wears motor and bearings. Worn out bearings create noise and heat. The heads are moving. And over time material in general transforms. My rule of thumb is that when a rotating hard drive is 3 years old you should replace it if it has run 24/7.

SSDs have silicon chips that will eventually fail. It has voltage regulators that fail. Capacitors fail. It all fails. The cooler it runs (within limits) the longer it lasts.

I absolutely choose SSD over HDD in my Home Assistant machine.

But even with SSD I ensure that

my SSD is much larger than the data stored on it
not leaving large unused gigabit files from forgotten VMs or containers from experiments
limit the recorder function in Home Assistant to the absolute minimum. Why record data I never look at just because I can?
limit logging to what is needed. If you enable debug logging, remember to disable it again.
if the SSD is an M.2 type buy a cheap heatsink for it. They cost a few dollars on eBay. They are easy to mount with a cooling pad and an elastic band that comes with the heatsink.

And again, if a disk SD card, SSD, or HDD fails because of heat, replace it!

Valentino_Stillhardt · September 20, 2020, 6:42pm

Thanks for this excelent and extensive answer!

As I said, I already ordered a new mSATA SSD and I’ll replace it and follow your recommendations!

Thanks a lot.