[Modding][HowTo] Investigating Homey’s eMMC wear estimation

It was fixed :sweat_smile:

New stats:

 11:49:47 up 57 min,  1 user,  load average: 0.24, 0.20, 0.13
mmcblk0 425.528 M

…yep, still writes are higher than expected but in my case it has been also reduced significantly. Would be still great to see even more improvements

The question is how they did it.

1 Like

Any downsides to this experimental update? I am on stable currently and trying to figure out if I should stay on stable version or switch to experimental.

Why would you want to switch to experimental?

It depends. It’s experimental, but - eg. for me it works fine. But there are always risks.
In case you have bigger smart house, eg. more then 40 apps, I would recommend switching it asap with prior backup naturally.

I don’t but also I want to minimize the writes on the nand. Don’t want ending up with a brick if the nand goes bad :slight_smile: If they will push at least this issue to the next stable soon then maybe it’s worth waiting

I didn’t win.

1 Like

kodos for the team for pushing it so fast

Only took about a year and investigation from community members :wink:

3 Likes

Mostly a @Sharkys

3 Likes

Question: Do Free Storage move up anddown when having lots of writes?

Because my Free Storage insight isn’t moving much at all, so is that a sign of low amount of writes?

No, most writes are done to update existing data.

Interesting, with the latest stable release; the cpu spikes every 1-2 hours are also gone. Maybe that there is a connection between de reduction of data that is written every hour and the strange cpu load in the previous firmware versions. Anyway, the lifetime of the eMMC is now about 8x115 days (it it is a new eMMC of course)? Hopefully with a CM4108032 it is even longer due to load balancing. Already replaced it twice :see_no_evil:

1 Like

Ranney, I wonder, how bich is your smart house, eg. devices, flows etc? I consider mime as the bigger one while I’m on 30% remaining lifespan (estimated)

Haha, not very big I think

Let’s see;
Around 325 devices at the moment
Spread over zigbee, z-wave, LAN, Cloud and BLE (trying to get matter over thread working without succes)
Most intense app: Philips hue with 4 bridges

Around 350 flows. But some are quite big. (My biggest advanced flows do not fit in one but are a combination of multiple flows to make them ‘working’ :see_no_evil:)

Around 85 apps

O, and ~7500 log entries every day (Syslog)

So, not really extreme but more a standard Domotica user that is building for some years?

1 Like

But from devices point view and apps point of view, bigger then me - 285 devices / 66 apps, also I guess experiment “Power User” enabled. As, if the life expectation interpretation from mmc-util is right, I’m on ~75% passed, no wonder that in your case it already died ;-(

Hmm. Ok. Wel, maybe that I’m using my homey more intensive than I see it myself. I really like the product and what it can bring (supporting lots of different brands, powerful for me and ‘easy to use’ for the rest of the family.

I’m on my 3rd CM4 module. I must say, it is not a problem to have a hobby that cost some money. It is also not a big problem to need some spare devices. But now I have 2 CM4 modules and a complete homey pro (incl Ethernet module) on spare. And the issue, writing too much info. Hopefully they can reduce the amount of written data more and… hopefully there are not a big group of users in the near future with failing CM4/Homey.

Just checked some other stats. Homey is running here around 20 flows every second. Also big and compex flows. But the cpu load is low. So in fact it is not a big issue (as far as I can see). Possibly that running flows also mean that data is written to the eMMC?

Would be great if the eMMC wear level is also visible in the developer screen. Maybe that they are collecting this info on the background now?

Only if you want your customer service to be flooded by questions about this :joy:

I has been already reduced quite significantly…

Well, that a little too much, but Homey can handle that - I would never run so many flows so often :-))

No really - the issue was primarily the amount of RRD data (insights/graph), for Power users and saving them to storage immediately without any Caching. More apps more writes, almost growing exponentially … There are still some optimisations they can do, but that will not bring such a significant saving anymore.

Yes, they do

1 Like