Memory leak in Turris system

The same

What catch my eye is
On old Turris i see about 90% free RAM anytime(luci) Its older Turris cache hide from luci page?

I believe old turris is also based on Openwrt.

1 Like

Those spikes on the graph are very strange. Turris team should take a look at this issue.

Please see

On opkg segfaults. We know about it and it’s long standing issue. Opkg is just crappy peace of software and we don’t have capacity to fix every problem. For some reason with some combinations of software if frees some memory it needs later on. We fixed some problems and workaround others that it at least doesn’t breaks the system (as it was previously). But better fixes must come from mainstream. For now you might try to just specify packages you want to file /etc/updater/user.lua and start I know it isn’t perfect. But it’s what it is.

Back on problems with ram. Please check /proc/meminfo and field MemAvailable. That’s what you want to see. For some reason luci is using field MemFree. If you have for example drives connected, then cache creeps. And on comparison with old turris. Do you have any storage connected to it? Also it has different kernel (older one) and different kernel configuration.

On those spikes in graph. There are no real used memory spikes I see. Those spies are just in cache and drop can mean things like disconnected drive or remounted drive. In reality, just don’t care about cache. It does it’s thing and you don’t have to be worried about it.


btw. jow put a lot of work into opkg recently: Git - project/opkg-lede.git/shortlog maybe it’s worth a try

1 Like

Thank you for pointing that out. I am adding it to todo list.


thank you cynerd. I personally using external HDD connected to USB. Lets see if original Turris owner has something too. If not, this could be difference which I am searching…

Regarding Luci: I am using simple memory for collectd, I dont think it could be configured other way. It doesnt bother me as I know about this cache, but the message from original Turris owner was worth checking for me .)

Thanks for explaining some of the issues. Is the memory usage metric something you guys collect from the devices? I’d be useful for you to see what the real issues are after each update.

No we don’t collect device status. Data collection is for network analysis not for support. And minimum number of people has direct access to raw data. So this is information you have to query on your own if you want to send it on support.