Memory almost 100 % full, despite free space on the 8 GB external memory card from Phönix Contact

PLCnext Technology & PLCnext Controls
1

Hi all, i would like to ask why the cockpit in Engineer shows me that the memory is 96% full, even though an 8 GB SD card is inserted. Attached is the picture from the Engineer and the picture of the SSH connection with the command df -h. kind regards nilsbaurx

2

Hello nilsbaurx,

The Memory mentioned is the RAM of your device not the storage space.
So It seems like you are running a pretty large application or might have a memoryleak in some part of your C++ or unsafe C# code. You can check by rebooting and following the memory during the execution to check for continuous increases.

If you are online via a shell you can use „htop“ to see the memory usage.

Edit:
I just now saw in your „df -h“ print that you have balena running.
Please make sure to use lightwight containers that are optimized for embedded systems.
Running a fullfledged OS with only 500MB of RAM will drive the plc pretty quickly to its limit.

Also it might be a temporary high memory usage if you are currently installing balena or setting up new containers. Do you have any problem with your application or are you just curious what causes the huge memory usage?
kind regards,
Oliver

3

Hi Oliver, first of all, thank you for your quick reply. Then the translation in Engineer is not really correct, because under memory I do not understand the working memory, but that is a matter of interpretation. Thanks for the htop command, I personally didn’t know it at all. I have attached a screenshot for you. The RAM is permanently occupied to approx. 65 %, but I think it is still within the green range. We have set up 3 Docker containers:

Node-RED
influxDB
Grafana Everything seems to run quite stable as long as the Grafana dashboard doesn’t get too big, for our use case we only have two graph panels, which run quite ok. However, it has happened that we wanted to display the complete last week in Grafana, for example, with 9 measured values with one data point every 5 seconds there are quite a lot of data points, so that it has also happened that the runtime on PLCnext has completely crashed. The AXC F 2152 with 500MB RAM is probably not the best controller for our purpose, but in the end it’s just a final project, and all the things that have to work also run quite well. For example, we also run a retention policy on our influxDB so that we automatically delete all data older than 24 hours so as not to unnecessarily burden the storage space on the PLCnext. We save the data every 24 hours via Node-RED to an FTP server.
In the end, we can live with this. Many greetings
nilsbaurx

htop.PNG

4

You can also clearly see in htop that as soon as I want to display the last 2 days in Grafana, for example, the CPU load goes up to 100%. We have just calculated how many data points Grafana has to pull out of the Influx: 9x12x60x24x2 = 311,040 data points. That’s a lot, probably too much for the AXC F 2152.

5

Hello nils,

For Memory usage the key is:

                                             * **Green** : Used memory pages
                                             * **Blue** : Buffer pages
                                             * **Yellow** : Cache pages

So with buffer and cache you are basically at 100% memory usage. Well I guess the CPU load is just temporary at 100% while generating the sheet?
As long as the CPU load falls back lower after that the RT kernel should guarantee correct behavior even in that szenario.

I guess memory should have been translated to „Arbeitsspeicher“ instead, you are correct.
I will forward this as an improvement request thank you for mentioning it. kind regards,
Oliver