Can I disable gl_monitor, gl_health?

souravndp · 2018-11-03T06:16:03.932Z

kyson-lok:

We’re working on improving the performance of MIIF, it is almost overload now.

Thank you for your clarification. I use cable internet, thus I assume that I can safely disable those. I am using GL-AR150. The main memory is being consumed by admin panel. lighttpd is more memory intensive than uhttpd. Moreover you have another www/api binary which is also taking a lot of memory. Have you thought about using websockets rather than HTTP requests in the background? there is a Github repo called ‘lua-websocket’, I have used it on openWRT for my own hobby project to display realtime graph and syslog.

Also, I don’t know if this is due to openWRT upgrade or not, v2.17(using LEDE) used to have about 0.1 1-minute load average when the router was almost idle. But now it is always more than 1.0 and sometimes it goes around 5. eg: currently, it is: 5.07, 3.80, 2.31

kennethrc:

… are you running out of memory somehow?

Not currently after I disabled unbound. It was very memory intensive, taking about 4MB memory. In v2.27, I used to have about 19-20MB free total memory (free+buffered) but in v3.0.5, I have about 13-14MB.

kennethrc · 2018-11-03T06:40:48.762Z

Free memory is useless memory. That’s nothing to be concerned about.

souravndp · 2018-11-03T06:55:43.702Z

kennethrc:

Free memory is useless memory. That’s nothing to be concerned about.

I understand that. But a fresh install with factory default settings should not have only 10-15 MB free memory. Because I have my own packages to install. eg: collectd statistics, some of my own maintenance scripts. From my previous experience, if memory drops too low, the router restarts.

And this is only due to the new admin panel. I know that it very useful than before but it is memory intensive. openWRT started using luci (which uses very lightweight lua language) to optimise their performance.

kennethrc · 2018-11-03T12:40:25.858Z

souravndp:

But a fresh install with factory default settings should not have only 10-15 MB free memory.

Why not? Again, “free memory is wasted memory”. What does “top” on the device say? Where is the used memory allocated?

souravndp · 2018-11-03T16:07:24.130Z

kennethrc:

Where is the used memory allocated?

lighttpd 7%
www/api 6-7%
www/html 4%

nmbd, smbd 4%

gl_health 3%
gl_monitor 2%

hostapd, procd, rpcd, netifd, odhcpd, logd, ntpd, ubusd etc system processes are taking 2-3% each.

The top 3 memory consumptions are due to the admin panel. That’s why I was requesting to optimize it. I love the new admin panel. especially the VPN configuration UI and the software installation UI because those things don’t have such great UI in the luci interface.

kennethrc · 2018-11-03T18:49:19.593Z

No, those aren’t telling us the right information. ssh into the device and run “cat /proc/meminfo”.

if most of your “used memory” is merely “buffers” or “cache” then it’s doing what it’s supposed to, and like I keep saying nothing to worry about.

souravndp · 2018-11-03T19:18:31.871Z

MemTotal: 59912 kB
MemFree: 9876 kB
MemAvailable: 23148 kB
Buffers: 6272 kB
Cached: 17604 kB

These are confusing. Different commands show the outputs differently. free shows some other outputs. The top command output and the luci web interface output matches. How ever as I read in some article, MemAvailable is the actual memory available for newly started program without swapping.

Johnex · 2018-11-03T19:37:45.816Z

MemFree: The amount of physical RAM, in kilobytes, left unused by the system.

MemAvailable: An estimate of how much memory is available for starting new applications, without swapping. Calculated from MemFree, SReclaimable, the size of the file LRU lists, and the low watermarks in each zone. The estimate takes into account that the system needs some page cache to function well, and that not all reclaimable slab will be reclaimable, due to items being in use. The impact of those factors will vary from system to system.

kennethrc · 2018-11-03T19:45:17.266Z

I wish you’d included the full meminfo output (namely, the “anonymous” entries), but:

souravndp:

MemTotal: 59912 kB
MemFree: 9876 kB
MemAvailable: 23148 kB
Buffers: 6272 kB
Cached: 17604 kB

“Cached” and “Buffers” hold data and/or programs in memory, so they don’t have to be pulled from the far-slower flash memory, and most of that can be reclaimed if necessary. Not only that, inactive processes will have their memory reclaimed, and paged back in if they’re needed. Not only that, many programs link to shared libraries, which (depending on the tool) “contribute” to the size of a program, but they’re really only in memory one time, whether one program is using it, or 20.

As I’d said before, you have plenty of memory left over for anything reasonable you’d like to do with this box, stop worrying about it.

Johnex · 2018-11-03T19:49:03.494Z

Yup, he would probably have to optimize his programs if they really need 20mb or more ram on a router…

souravndp · 2018-11-03T19:56:45.433Z

Thank you @kennethrc and @Johnex for your clarification.

souravndp · 2018-11-04T13:47:21.220Z

Hi, @Johnex @kennethrc Just a quick question.

root@MrGhosh:~# cat /proc/loadavg
1.37 1.09 1.08 1/64 29235

I read that for a single core CPU, load avg of 1 means that the processor was optimally used. No process was waiting for CPU time. In my v2.27 firmware, the load averages were used to be around 0 (say, 0.11, 0.30 etc. because the system was mostly idle), but now, it is always more than 1 (the system is factory default and by always I mean it never comes below 1) and sometimes it goes around 3-4. Should I be worried?

Johnex · 2018-11-04T15:28:08.521Z

You can look in Luci at the process list, check which processes are using most CPU. Ofc that while using the UI, or having it open, that is a lot of load on the cpu, the GL UI is doing many requests per second to update all the data.

souravndp · 2018-11-04T16:13:43.533Z

Johnex:

check which processes are using most CPU

Please have a look at the following screenshot. No process is taking much of CPU. Still, the load average never drops below 1. I am not using any UI (GL admin panel or luci) currently. Still such high load.

I was concerned about memory and you guys explained that it was fine. But I think that I should be worried about the CPU load. Should I open a new thread about this?

Moreover, I could not sort the results. This is busybox top and original top command’s switches are not working. How to sort the processes based on memory,cpu etc?

image1366×768 96.1 KB

Johnex · 2018-11-04T16:19:44.237Z

Well you can see yourself. Just 1 process is sitting there working, which is top. The other processes are idle.
Look a the cpu usage not the “load averages”.

souravndp · 2018-11-04T16:28:58.101Z

Johnex:

Look a the cpu usage not the “load averages”.

Load average values are a very standard way to check system load. Those values don’t lie.

scoutapm.com

Understanding Linux CPU Load - when should you be worried? | Scout APM Blog

In terms of load averages the three numbers represent averages over progressively longer periods of time (i.e. 1, 5, and 15-min. averages).

kennethrc · 2018-11-04T16:33:40.336Z

… talk about “making something out of nothing”. Good luck, I don’t know what you think you’re seeing. though.

souravndp · 2018-11-04T16:41:02.145Z

kennethrc:

… talk about “making something out of nothing”. Good luck, I don’t know what you think you’re seeing. though.

A load average never going below 1 and you claim that this is fine? load average has no relation with CPU usage. this is the measure of load on the system.

I understand that I was over-concerned about the RAM issue but as far as I understand, this load average is not normal. My usage pattern has not changed and in firmware v2.27, My load wad around 0.1 generally

Johnex · 2018-11-04T16:49:50.213Z

You are the only user that has an issue with this. If you feel that you need more power, get a B1300 that is quad core and will show probably 0 at all times. Unless you see actual issues when running your programs, the stats don’t matter honestly.

Do real world testing with your programs.

kennethrc · 2018-11-04T17:11:59.104Z

souravndp:

A load average never going below 1 and you claim that this is fine?

OMG … “load average” is “average number of processes in ‘ready to run’ state” over some period of time.

… so question: what is marshalling out the output of “top” from the stdout buffers, to the socket, thru the networking stack, and out the interface? Invisible fairies, or the CPU, which has to be running a thread or process to do all that?

John is right- if you have this many doubts (which is “inexperience”, IMO) about the GL-iNet devices (which are highly rated, I have three of various models) then yeah, you should get another brand.