Merlin 380.57 Potential Memory Issue

[Nullity]

I think we need to pay much less attention to total RAM usage, and more attention to the "Active" RAM, as reported by "cat /proc/meminfo", because cached memory usage (the majority of RAM usage) tells us nothing about the active, vital memory used by programs.

My RT-N66U running john's fork with a 4 day uptime:

admin@merlin:/tmp/home/root# grep -E 'Active|Cached' /proc/meminfo
Cached:          15548 kB
SwapCached:          0 kB
Active:           5716 kB

 

[MarkyPancake]

Isn't the issue that RAM used has gone up in more recent builds and didn't happen to the extent people are noticing it now in older builds.

 

[Nullity]

Isn't the issue that RAM used has gone up in more recent builds and didn't happen to the extent people are noticing it now in older builds.

Is it an issue? RAM usage itself (including cached) tells us nothing. Heck, higher usage could even be because the memory management decided to cache more intelligently.

Too many bored people see some graph a simply want to "fix" it if it feels high. I do it...

 

[sfx2000]

From what I'm seeing cruising around the Asus forums here - the 380 upstream seems be a bit of a challenge, quite a few issues noted outside of RMerlin's changes...

 

[jegesq]

From what I'm seeing cruising around the Asus forums here - the 380 upstream seems be a bit of a challenge, quite a few issues noted outside of RMerlin's changes...

Honestly, other than observing the RAM usage I described initially after upgrading to 380.57, I've seen absolutely nothing in the way of adverse behavior from my AC3200. Upstream and down are fine, nice and snappy. What "upstream" issues are you reading about with this FW. The only issues I've read about are a loss of 2.4ghz and that's mostly been confined to the AC68's (and perhaps AC87's).

I do agree that perhaps the title of this thread should have been followed by a "?" because the way it reads now, I think a lot of people have assumed that I was trying to describe something that was an "actual" (rather than potential") memory issue. Indeed, as it turned out, there was no issue at all, and a reboot seems to have cleared out whatever was running that was consuming memory (and of course as Nullity notes, it could also be that higher committed used RAM is just an indication that the kernel is taking on a larger management role with respect to things that it might not have been as effectively managing in older versions....I don't know whether that's the case or not, but I certainly get his point).

 

[sfx2000]

What "upstream" issues are you reading about with this FW. The only issues I've read about are a loss of 2.4ghz and that's mostly been confined to the AC68's (and perhaps AC87's).

It's not just RMerlin's fork that has issues - even the factory build seems to have them...

When I speak about upstream/downstream is what the sources are - in this case, Asus's GPL drops...

I don't think RMerlin's changes have any impact here with the "380" problems...

 

[Nullity]

Hehe. sfx2000 means "upstream" as in AsusWRT is upstream of AsusWRT-merlin, just like FreeBSD is upstream of pfSense.

Developer code-base jargon.

 

[jegesq]

Hehe. sfx2000 means "upstream" as in AsusWRT is upstream of AsusWRT-merlin, just like FreeBSD is upstream of pfSense.

Developer code-base jargon.

Got it. Not a dev, so it went right over my head. Thanks.

 

[MarkyPancake]

My used RAM had gone up to 81 MB today, then after I made a change to the 2.4 GHz network and applied it, the used RAM dropped down to 45 MB.

 

[eddiez]

My used RAM had gone up to 81 MB today, then after I made a change to the 2.4 GHz network and applied it, the used RAM dropped down to 45 MB.

Restart of service, so makes sense. Cached data cleared by restart.

 

[sfx2000]

Developer code-base jargon

Sorry folks - FWIW - lot's of smart folks on the board, sometimes I forget not everyone is a dev... (and I'm not a Dev on any of the AsusWRT builds, can't right now for various reasons...)

 

[Nullity]

Sorry folks - FWIW - lot's of smart folks on the board, sometimes I forget not everyone is a dev... (and I'm not a Dev on any of the AsusWRT builds, can't right now for various reasons...)

It impresses me that so many "normal" users can quickly learn Unix-based systems like AsusWRT when I, even after over a decade of using Unix-based systems as my day-to-day OS, have trouble understanding many aspects of AsusWRT.

 

[GreGHM]

It impresses me that so many "normal" users can quickly learn Unix-based systems like AsusWRT when I, even after over a decade of using Unix-based systems as my day-to-day OS, have trouble understanding many aspects of AsusWRT.

That's an Internet power!
Now each person with laptop and Google in 10 minutes becomes a qualified surgery specialist-doctor with clear understanding that right hand should be cutted off...DONE! oh no.. Left hand! DONE!.... Oh NO! Left foot! DONE! ))

 

[Nullity]

That's an Internet power!
Now each person with laptop and Google in 10 minutes becomes a qualified surgery specialist-doctor with clear understanding that right hand should be cutted off...DONE! oh no.. Left hand! DONE!.... Oh NO! Left foot! DONE! ))

By "laptop" I think you mean "watch" or "phone". I am both happy & sad to feel so old at 30.

Technology!

PS - Surgery is as easy as running (QWOP).

 

[MarkyPancake]

My AC68U's RAM used is up to around 100 MB now, but what I don't understand about this is when the router's settings and features remain constant and the daily use across all the connected devices is the same behaviour, why does this use keep creeping up? It's not like more devices are connected or new services are being used internally or externally.

 

[highwire]

My take is this. If the router is functioning normally, who cares how much memory it's using? I'm as OCD as anyone, but some of us need to stop obsessing over stuff that just doesn't matter.

 

[MarkyPancake]

My RAM has gone from 94 MB to 106 MB when no one has been using our broadband today. How is that normal and what happens if it does get to the point where it uses it all.

In comparison, my N66U saw months of uptime, performed great and the RAM use stayed constant serving the same connection and devices using John's fork.

Just seems unusual, especially as my AC68U isn't doing anything over what the N66U did and neither did or are doing anything more than basic router functions. Nothing funky is being used like VPN or any custom scripts, etc.

 

[Nullity]

My RAM has gone from 94 MB to 106 MB when no one has been using our broadband today. How is that normal and what happens if it does get to the point where it uses it all.

In comparison, my N66U saw months of uptime, performed great and the RAM use stayed constant serving the same connection and devices using John's fork.

Just seems unusual, especially as my AC68U isn't doing anything over what the N66U did and neither did or are doing anything more than basic router functions. Nothing funky is being used like VPN or any custom scripts, etc.

94 to 106 is a few percent change. Insignificant.

You ask "how is that normal?", but I think the first question you should ask is "is this abnormal?". Do you have enough information to deduce that increased RAM usage is bad?

Until there is no RAM free, there is not even the beginnings of a problem. Even when all RAM is used, most of it is cached data, which is non-vital and can be cleared whenever a program requires more "active" memory.



We need to prove that this increase in mem usage is a problem before we go searching for a solution. Premature optimization is a major sin of computing.

 

[eddiez]

My RAM has gone from 94 MB to 106 MB when no one has been using our broadband today. How is that normal and what happens if it does get to the point where it uses it all.

In comparison, my N66U saw months of uptime, performed great and the RAM use stayed constant serving the same connection and devices using John's fork.

Just seems unusual, especially as my AC68U isn't doing anything over what the N66U did and neither did or are doing anything more than basic router functions. Nothing funky is being used like VPN or any custom scripts, etc.

Even though 'no one is using broadband', you are still connected to internet with your router. So it's not that everything is idle. Remember that a running engine still uses gasoline, even when it remains stationary...

 

[john9527]

In comparison, my N66U saw months of uptime, performed great and the RAM use stayed constant serving the same connection and devices using John's fork.

Since you made the comparison, a couple possibilities come to mind......
- The later Merlin networkmap runs more frequently and collects more information than my fork. As clients come and go, it may be saving some additional information in buffers until the space is needed.
- The later Merlin also includes the TrendMicro DPI engine when running on the AC68U. If your are running any of the DPI functions, it could easily be collecting/buffering information. I also personally think that parts of this code may be running as an 'assist', such as with CTF, even if you aren't explicitly using the DPI functions possibly contributing to some addition memory use.