100% CPU usage on TPLink WR1043 v1

[elijahg]

Hey guys,

For some strange reason my TPLink WR1043 with LEDE 17.01.2 occasionally goes into overdrive, and the CPU sits at 100% with 13 load average, according to htop. Pings are lost or severely delayed (200ms+) and web browsing slows to a crawl. Nothing in particular seems to be using excessive CPU, so I'm not sure what's going on. I can't access LUCI with the error "Bad gateway". It happens randomly (after 4 days this time) and a reboot fixes it. I've attached a screenshot of htop, if anyone has any ideas that'd be great.

Thanks!

[gwlim]

Try my fastpath build optimized for mips24k 32MB

gwlim/Fast-Path-LEDE-OpenWRT

[elijahg]

Many thanks gwlim, your build does seem to have fixed it so far! If it dies again I'll update this thread

Funny enough another router on the network also on LEDE has just self-rebooted after similar symptoms, gonna keep an eye on that one...

[weedy]

Living with a TL-WR842ND (sysupgrade bricks router)

I think there might be a bug somewhere. I’ve just noticed that sys% usage is ridiculous when everything goes to hell. >pppoe-wan │ 494.90KiB 351 pps │ 12.13KiB 192 pps Mem: 19620K used, 8308K free, 4K shrd, 864K buff, 1472K cached CPU: 1% usr 93% sys 0% nic 0% idle 0% io 0% irq 4% sirq Load average: 6.89 4.88 2.96 5/58 9906 PID PPID USER STAT VSZ %VSZ %CPU COMMAND 1555 1 root R 1948 7% 16% /usr/sbin/hostapd -s -P /var/run/wi…

If it happens again look to see what your sys usage is.

[elijahg]

Unfortunately the same seems to still be happening. Sys usage is 100%, but adding up the running processes CPU usage only adds up to around 10%. This happens any time I try to access the web interface.

[tmomas]

Which version of TL-WR1043ND are we talking about? v1 perhaps?

[elijahg]

It is v1, somehow missed adding that to the title... Oops.

[tmomas]

TL-WR1043ND v1 has only 32MB RAM.

Devices with ≤4MB flash and/or ≤32MB ram suffer from limitations in extensibility and stability of operation.

See https://lede-project.org/meta/infobox/432_warning for details.

See your logfiles if you can find something usefull.

[weedy]

Except my TL-WR842ND has a few MiB left of ram on top of having another 2MiB in caches that could be dropped. The 100%sys doesn't make sense.

Mem: 19620K used, 8308K free, 4K shrd, 864K buff, 1472K cached
Mem: 21212K used, 6716K free, 12K shrd, 792K buff, 1584K cached

[elijahg]

I realise 32MB RAM is quite limiting, but still as Weedy points out below the 100% CPU usage was weird, it was happening for me even with several megabytes free. There was nothing in the log files of any use unfortunately.

I've since built my own image and for the last couple of days things seems to have been stable, with load averages of 0.01 or less. I've got about 5MB RAM free, also stable over the last few days.

[raz123]

I’ve since built my own image and for the last couple of days things seems to have been stable, with load averages of 0.01 or less. I’ve got about 5MB RAM free, also stable over the last few days.

Care to share?

[r43k3n]

I also had 100% CPU usage with WR1043NDv1.
I think it's caused by very low RAM memory available.

[cbz]

I have a similar issue on the W8970, except during these spikes all connections will also be dropped - though the ADSL connection itself will stay up. It's as if it loses all NAT mappings at this points.

The spikes are generally shorter (and the W8970 has 64MB of memory)

[elijahg]

Care to share?

Here you go, it's got Luci and IPv6 support built-in

Firmware

I have a similar issue on the W8970, except during these spikes all connections will also be dropped - though the ADSL connection itself will stay up. It’s as if it loses all NAT mappings at this points.

The spikes are generally shorter (and the W8970 has 64MB of memory)

The PPPoE session on my WR1043v1 stays up, but it's just incredibly slow and it does drop a lot of packets.

That said, I'm on 2 days uptime now and it's still going strong, load averages of 0.01 and about 3MB free RAM. I'm going to enable swap space on a USB stick, will be interesting to see if the kernel actually uses any. If it does then it's definitely running out of RAM.

[raz123]

I’m going to enable swap space on a USB stick, will be interesting to see if the kernel actually uses any. If it does then it’s definitely running out of RAM.

Let us know how you fare! Oh, and by the way, did you end up using dissent1's Fast Path (Shortcut Forward Engine) patch? (https://github.com/lede-project/source/pull/1269)

[elijahg]

Will do! I did give it a go, but unfortunately had the same problem after a day or so. If my Github skills were up to scratch I'd merge dissent1's changes into my build, but alas they are lacking!