RUTX50 Recurring OOM events and excessive slab_unreclaimable growth

Networking Solutions Issue
1

Device: RUTX50
Firmware: RutOS 00.07.22.3

Description

I am opening this forum thread following a discussion with Teltonika support, who asked me to share my findings here for further investigation. @Justinas

I am experiencing recurring Out-Of-Memory (OOM) events on a RUTX50 running RutOS 00.07.22.3. The router usually operates normally for around 24 to 72 hours before becoming partially or completely unresponsive. In most cases management access is lost and network connectivity is disrupted until the device is rebooted.

Configuration

My setup is relatively simple:

  • Mobile network as the primary WAN connection

  • WiFi WAN configured as a backup connection

  • WiFi access points for client devices

  • WireGuard tunnel used as the default route for internet traffic

  • No additional packages installed

  • No custom services or scripts running

What I investigated

My first assumption was that connection tracking exhaustion or a userspace process was consuming memory.

However, connection tracking usage remained very low:

nf_conntrack_count: 104
nf_conntrack_max: 31744

I also reviewed the OOM logs and noticed that the kernel was terminating a variety of unrelated processes, including:

  • hostapd

  • wpa_supplicant

  • mobifd

  • rpcd

  • event_server

  • mdcollectd

Example log entries:

Out of memory: Killed process 4120 (hostapd)
Out of memory: Killed process 4811 (mobifd)
Out of memory: Killed process 4467 (rpcd)

The affected processes all had very small RSS values, which made me suspect that they were victims of a wider memory issue rather than the actual source of the problem.

Findings from the troubleshooting package:

While reviewing the troubleshooting archive, I found the following memory information during one of the OOM events:

slab_reclaimable: 977 pages
slab_unreclaimable: 45505 pages

Based on a 4 KB page size, this corresponds to approximately 178 MB of unreclaimable slab memory on a device with about 246 MB of total RAM.

This suggests that the majority of memory consumption was occurring inside the kernel rather than in userspace.

Additional observations

Around the same timeframe, I also found repeated wireless driver messages such as:

ath10k_ahb … could not get mac80211 beacon

and

ath10k_ahb … peer-unmap-event: unknown peer id

Potential trigger

One pattern I noticed is that several previous incidents occurred after I manually switched from the mobile connection to the WiFi WAN backup due to poor cellular reception and disabled the modem manually.

I cannot say with certainty that this is the trigger, but it may be relevant because several OOM-related log entries reference mobile networking components such as:

  • mobifd

  • uqmi

  • dhcp_mobile.scr

At the moment my best guess is that there may be a kernel-level resource leak involving the wireless and/or mobile networking stack, possibly related to interface transitions.

Questions

  • Has anyone observed similar OOM behaviour on RutOS 00.07.22.3?

  • Are there any known issues involving slab memory growth, ath10k, WiFi WAN, or mobile networking components?

  • Are there any additional diagnostics I should collect if the issue occurs again?

I still have the troubleshooting archives and can provide additional logs if required.

Thank you.