5.0rc1 Profiler unclassified CPU usage

Seems there is an unclassified process eating cpu.
No problem in RB411ah but on older mipsle’s:

[admin@a8] > system resource print
uptime: 1h38m21s
version: “5.0rc1”
free-memory: 17100KiB
total-memory: 29612KiB
cpu: “MIPS 4Kc V0.10”
cpu-count: 1
cpu-frequency: 266MHz
cpu-load: 8%
free-hdd-space: 90112KiB
total-hdd-space: 126976KiB
write-sect-since-reboot: 75
write-sect-total: 365124
bad-blocks: 0.3%
architecture-name: “mipsle”
board-name: “RB532”
platform: “MikroTik”
[admin@a8] > tool profile duration=10s
NAME USAGE
wireless 2.5%
console 0.5%
winbox 1.5%
logging 1%
management 2%
profiling 8.5%
unclassified 84%




[[admin@c999] > system resource print
uptime: 10m9s
version: “5.0rc1”
free-memory: 4108KiB
total-memory: 13368KiB
cpu: “MIPS 4Kc V0.11”
cpu-count: 1
cpu-frequency: 175MHz
cpu-load: 27%
free-hdd-space: 35212KiB
total-hdd-space: 61440KiB
write-sect-since-reboot: 214
write-sect-total: 620070
bad-blocks: 0%
architecture-name: “mipsle”
board-name: “RB133C3”
platform: “MikroTik”
[admin@c999] > tool profile duration=10s
NAME USAGE
console 0.5%
flash 0%
winbox 1%
logging 2%
management 4%
profiling 14%
unclassified 78.5%

if unclassified is larger than 5%, make supout.rif file and email support, we will then give name to the unknown process for the next version

Sent a supout. I hope the cpu hog gets a name and is corrected to use
less cpu. The systems are idle.

NOTE: RB133/RB532 doesn’t work with the profiler correctly, we won’t be able to fix that. Most of your unclassified process is IDLE. The RB133 series doesn’t detect IDLE correctly.

I sent a supout.rif of the RB500. The RB133 dies while generating a supout.
I had to reboot it to access it again.

5.0rc1 is usable on RB500. RB133 is very slow (dont expect it to perform like hell but would like
to keep them running for users with <5MBit/s usage).

many people sent us supout.rif files from RB133, so it’s possible that you have some heavy configuration on it ?

Ok, also RB532 is not supported entirely. the IDLE is included in Unclassified.

MIPS-LE systems are not completely supported

[admin@MikroTik] > tool profile duration=10s
NAME USAGE
wireless 5%
snmp 13%
ethernet 4%
console 0%
flash 0%
firewall 0%
winbox 0%
logging 2%
management 45.5%
idle 18%
profiling 0%
queuing 2%
bridging 8%
ppp 0%
unclassified 2.5%





management 45.5% - ??? RB-800 5 RC 1

NAME VERSION SCHEDULED

0 system 5.0rc1
1 routeros-powerpc 5.0rc1
2 X dude 4.0beta1
3 X dhcp 5.0rc1
4 X ipv6 5.0rc1
5 X hotspot 5.0rc1
6 wireless 5.0rc1
7 routerboard 5.0rc1
8 ppp 5.0rc1
9 routing 5.0rc1
10 X mpls 5.0rc1
11 advanced-tools 5.0rc1
12 security 5.0rc1

I was having massive problems with the routing task sucking up all the CPU on a RB433AH. I am using ospf and there seems to be some sort of connection there. I was unable to create a supout file as there was not enough cpu left to make the file in a reasonable timeframe.

I had to roll back to V4.11. A shame too as Nv2 looks promising.

I am using 802.11a cards. SR5 I think.

Hi,

What is the item “management” of profiler? Please post here, if any body know about it.

We have some CPU eating process which is classified as unknown and which is degrading our main routers performances:

ROS 5.13, CPU Q6600. We contacted official support regarding this 3 days ago, but we haven’t received any answer yet (except autoreply).

We’ve tried to reset-configuration and configur everything from scratch, but we still get this process. I am not sure about older versions of Mikrotik because we employed this router from ROS 5.11 and it behaves the same with 5.11, 5.12 and 5.13, unknown process remains.

We solved our problem, but I want to share this case with you. Today we had complete stop of this router so I went on spot and played a little with BIOS options. Original idea was to stop reminding options which are not necessary, but I turned CPU fan warning also and when I restarted the machine, I could hear the warning immediatelly. When we opened the case CPU fan was really not working because one wire blocked it. After we removed and fixed the wire properly, our router started normally and CPU usage was less than 10% per core. So, CPU freqency was lowered because of the CPU temperature, but we couldn’t debug this from /system resources because it displays default CPU frequency, not the running (scaled) one. We couldn’t trust to /system health either because there are no exact fan labels (CPU fan, PSU fan, etc.) and there are some strange readings like negative temperature.

huh… I don’t have anything in /sys health for x86 routers… what platform do you have?..