@yk said in MMM-OpenWeatherForecast - Replacement for MMM-DarkSkyForecast:

Fantastic work! I was using the previous DarkSkyForecast version - easy shift to this one.

I’m trying to put the hourly data and daily data in the same “row” across using tiled view instead of 2 rows. I’m happy to edit the njk file (have tweaked a bit haven’t gotten it “just right” yet). Any ideas? I gather it’s something with the hourly and daily divs.

Is data available to get mm of precipitation in the hourly and daily blocks? It shows % chance - keen to add precipitation instead.

Thanks!!

Replying to my own post incase someone else wants a single row for 3 hours and 3 days. You’ll need to comment out lines 191, 309, 315, 318 in the njk file. This results in a single row readable left to right hours then days. It’s a useful view if you’re using a full width view.

Fantastic work! I was using the previous DarkSkyForecast version - easy shift to this one.

I’m trying to put the hourly data and daily data in the same “row” across using tiled view instead of 2 rows. I’m happy to edit the njk file (have tweaked a bit haven’t gotten it “just right” yet). Any ideas? I gather it’s something with the hourly and daily divs.

Is data available to get mm of precipitation in the hourly and daily blocks? It shows % chance - keen to add precipitation instead.

Thanks!!

Understood. Thanks.

Server seems doable. With the NUC it’ll have a bit more horsepower to run my iCloud sync and Weather setups (nothing overly painful to be honest).

Does the client 3B also run MM then? Meaning run your install scripts but instead update mm.sh to run client?

I fear I’m making the setup overly complex (only satisfying my tinkering needs). Would you agree or have you found that a server > client setup works better? (curious of opinion)

Thanks Sam. The 3B has been running fine for years. Now that the card failed I’m tempted to go the server > client route. So the NUC would run Debian, kick off your scripts, it runs in server mode, I do all of the normal setup that I did previously on the pi? How does the client setup work once I replace the failed card?

And, be honest, is this overkill? Once a new card comes in I can redo the setup on the 3B.

I’ve got a spare NUC. Fully appreciate that’s it’s a bit of overkill (I have my 3B running but a card failed). How would I go about getting the base setup going? Pm2 autoboot, base install, what OS?

Was there an agreed solution to this one? Do we need to subscribe separately to the 3.0 API or does the free subscription doable (works with default weather module)?

@sdetweil let’s close this one. There was a need to pass through a height or width value when using MMM-Unsplash. Passing nothing seems to have grabbed the raw file and chewed up a good deal of memory. I’m still monitoring.

I’d love to setup a log element that drops memory usage into the pm2 logs. Just fear monitoring memory may eat up too much memory ;) (but I should have plenty of it)

I’m using MMM-Tools to display stats, may leverage that.

@sdetweil I’m working with the developer of MMM-Unsplash. This might be down the the usage of “raw”. Will revert.

Separately, just to confirm: no issue with calendar fetcher broadcasting every 10 minutes? What is it doing? (can you tell I’m sniffing everything now?) ;)

@sdetweil agreed.

But i’m seeing the below after another memory issue today.

type total used free shared buff/cache availabl
Mem: 925 291 367 9 267 571
Swap: 1023 93 930

@sdetweil sorry

I bumped up the swap already to 1024

memory
total = 925
used = 272

swap
total=1023
used=116

@sdetweil keeps happening unfortunately. it might be time to start removing modules to see if that helps.

Two errors seen:
out of memory error and also saw a fatal: unable to access ‘https://github.com/MichMich/MagicMirror.git/’: gnutls_handshake() failed: Error in the push function.

I’m only using:
MMM-MyWastePickup (only uses local file)
MMM-Unsplash (hits Unsplash twice an hour)
MMM-DarkSkyForecast
MMM-Tools
Calendar does an iCloud refresh using vdirsyncer (every 30 minutes)

Calendar-Fetcher looks to be doing a broadcasting of events every 10 to 12 minutes. Not sure what this is.

mem total=925 / used=272
swap total=1023 / used=116

I wonder if the iCloud hit happens at the same time as the Unsplash hit causing an issue.

@sdetweil that’s a good point. It happened again - I gather that’s probably a best option.

Keep me honest, is the below the best approach?

For others searching in the future:

1. sudo dphys-swapfile swapoff
2. sudo nano /etc/dphys-swapfile
3. update CONF_SWAPSIZE to something like 1024
4. save file (ctrl + x)
5. sudo dphys-swapfile setup
6. sudo dphys-swapfile swapon
7. sudo reboot

You’re probably right, need to increase. Mine is set to 100 at the moment.

@sdetweil thanks!

Confirmed on the correct version

“version”: “2.15.0-develop”

Interesting item in the logs:

[975:0310/092506.157270:FATAL:memory.cc(22)] Out of memory. size=174170112

Only change would be the previous item you and I discussed around MMM-Unsplash. I’m curious if using the “raw” version of the image continues to chew up memory. Something to keep an eye on for sure. I’m using one of the tools modules to keep an eye on CPU and memory - I don’t see it growing throughout the day though. Perhaps it’s a “perfect storm of refreshes” that happens.

Next time I see the screen go blank I’m going to check logs immediately.

No changes to installation. Pi pointed to development branch. Recently screen going blank. A restart using pm2 fixes it. Going to start going through logs but it’s an odd one. Happy to do any other debugging the group feels would help (“can you check a,b,c”).

@sdetweil Fixed. It’s a change in the Unsplash API. obj.urls.custom is no longer an option returned in JSON. I changed “custom” to “raw” and it’s good.

Lovely.

@sdetweil hehe. Agreed. And now I have a decision to make. Hack it to work or rewrite. 😉

@sdetweil saw that. Love the fact that it worked just a few days ago. I have a funny feeling it’s between Unsplash API and new work in MM. I’m reading the API docs now. It’s a simple setup so hopefully a tweak shouldn’t be too tough.

Got it. Tweaked config.js to browse to ip:8080. I think you guys are right, there are going to be many modules broken. ;)

I’m going through normal debug type approaches to see what’s up. Seems Unsplash hasn’t been updated in ages. Can someone remind me how to inspect the rendered mm page remotely? I’m thinking I’ll just browse to the ip, and slowly go through each div via inspect. I’ve done it before but don’t remember how I did it. Painful but easily doable. This could just be a “something changed and it’s a style issue”.

I normally sit pointed towards the dev branch only because I’ve gotten used to helping out with calendar issues where I can. I noticed two days ago that a specific module (MMM-Unsplash) has stopped working. What’s the best way to check what’s changed to help troubleshoot the issue? If nothing has changed on my end (still same pi) then thinking it’s the dev branch change and perhaps may impact others?