Memory leaks?

[pz1]

I am only running v1.7.2 with HA where I have defined:

1. Battery polling,
2. Pol sensors periodically,
3. Logical rules

The logical rule is not functioning properly as I have reported in an other thread.
Because my system tends to crash, I am recording memory usage as requested by PoltoS. Below in the screendump are a couple of reports that show that in some 3 days, the used memory has almost doubled from 93064 to 169668.
My installation is based on NOOBS.

memory.PNG (326.71 KiB) Viewed 13598 times

Update: Now five days after the start used memory has risen to 226120. Half of the total.

[pz1]

The 6th day after the start of the above experiment, I did remove all automation modules in the preferences menu. Still as the image below shows, the amount of free memory continues to decrease, though it looks as if it is at a somewhat slower pace.

usage20140903-10.31.PNG (14.08 KiB) Viewed 13468 times

Only 203776 free now on day 8.

EDIT: 20140917: Since this post I did get the following two updates from Poltos via mail:

LEt me discuss this with pofs, but I don't see any critical things in your data: Look at RSS columnt. 22 Mb memory usage is pretty ok.

Does the overall free+cached memory goes down during time?
PS

looked a bit more. seems like it is not Z-Way eating all memory. And even not sure if it is - look at buffered + chached + free. It is almost constant. This is easy to check using some program that really eats memory - kernel at some point will decide to free buffers and cache to give it to the eater.

PS

I did not yet have the time for further experiments

[kickus]

It seems that I'm experiencing the same kind of memory leak.
I have a frequent polling of several electricity meters (5 devices, 17 values), and the z-way network gets unstable after a few days when the raspberry gets out of memory.

It appears as strange behavior. Some values does not get updated anymore (but others does). top command showed that z-way-server uses 65.9% memory and about 80-99% CPU. Free -h revealed that 418M was in use and 18M free.

After a z-way-server restart, the top command showed 4.4% memory usage, 10.2% CPU, and free -h 150M used, 286M free.

This was less than 5 days since reboot.

[pz1]

@kickus
I did update my post with some remarks I received later via email

[Lumberjack]

I do not have any issue with the CORE 1.7.2 release since 42 days of continuous uptime and very stable operation. The only thing I did was REMOVE all home automation which I do with my own python code. Before that, I also experienced performance degradation.

Since we are all posting tops, here is copy I made after 42 days (This particular Raspberry has 32 Z-wave devices connected):

top - 17:46:35 up 42 days, 6:32, 1 user, load average: 0.15, 0.18, 0.21
Tasks: 68 total, 1 running, 67 sleeping, 0 stopped, 0 zombie
%Cpu(s): 9.4 us, 3.0 sy, 0.0 ni, 87.6 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem: 447996 total, 435492 used, 12504 free, 77404 buffers
KiB Swap: 102396 total, 0 used, 102396 free, 231104 cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
2049 root 20 0 501m 55m 3248 S 5.9 12.7 2954:11 z-way-server

[dubliner]

I do not have any issue with the CORE 1.7.2 release since 42 days of continuous uptime and very stable operation. The only thing I did was REMOVE all home automation which I do with my own python code.

Can I ask how you trigger an update?
I've tried logging using HA but after a week I've an average of two temperature readings a day, probably from manual updates I did via the blue UI.
So now I'm trying this in a cron job:
wget "http://localhost:8083/ZWaveAPI/Run/devi ... 49].data[1]"
Is this the best format command for me? Does anyone know how I force an update of either the Fibaro Universal with temperature sensors, or a Topp 123658 power meter?

Any ideas much appreciated.

[kickus]

It seems to me that version 1.7.2 is still consuming quite much memory. Uptime: 1 day, 23:16

memory leaks with 1.7.2

z-wave-memory.PNG (32.44 KiB) Viewed 12483 times

[kickus]

I'm not using any automation at all. I do not do any polling.
I'm just running the controller function.

In about 22 hours, the z-way-server has consumed notably amount of the raspberry memory.
As earlier mentioned, I'm running version 1.7.2.

Is this a problem that everybody is phasing, or is it just me?

Memory wasted again

z-wave-memory2.PNG (20.74 KiB) Viewed 12360 times

...in about 22 hours

z-wave-memory3.PNG (4.01 KiB) Viewed 12360 times

[pz1]

Is this a problem that everybody is phasing, or is it just me?

I would rephrase that as: or is it just us?

I still haven't solved it. Please note that you have a few more processes running than I do. My installations are based on the standard virgin image, where I prevented openremote to run (see other thread on this forum). The latter has improved my situation a bit

[kickus]

Is this a problem that everybody is phasing, or is it just me?

I would rephrase that as: or is it just us?

Sorry pz1, thanks for pointing out that you're still suffering the same problem.
According to your signature, you're running Z-Way 2.0.0-rc8, and it seems to have the same troubles as 1.7.2.

My image is not much modified either. Compared to a virgin image, I do run a node js script that extracts interesting data from the z-way http interface into files on small ramdisk. I do also run a bash script that forwards the data to a centralized server (using curl). The purpose is to collect energy monitoring data from many similar raspberries (within an EU project). Both of my scripts are quite simple and light weight (both in functionality and footprint) compared to the z-way-server.

I had to restart my z-way-server process now, because the raspberry began to jam. I did also stop the openremote process in order to follow up whether it has any affect for me. After the process restart, the z-way ui is snappy and responsive again.

The raspberry has even started to swap to the sd card.