problem with v3.4

[david_c_1]

Hi,

I have installed version 3.4 but when I start the basic example i get the following warning in the log:

 WARN: ClusterManager - schedule failed, purpose: portal task: 76924429298 reason: No worker available, all in full load.

In the console of browser i get:

401 (Unauthorized) - invalid token

[Qiujiao_W_Intel]

According to the warning message:

WARN: ClusterManager - schedule failed, purpose: portal task: 76924429298 reason: No worker available, all in full load.

the device running portal agent is in full load, please double check whether the CPU usage on this device is too high 

[david_c_1]

I have installed the release 3.4 on three different machines (1 centos 7.2, 2 ubuntu 16.04). On these machines I always get the same error of full load.
For example, with woogeen stopped  on the CentOS machine I have the followinf cpu load average: 0.04 - 0.03 - 0.05. 
After starting woogeen I get the full load error with a cpu load of 0.55 - 0.33 - 0.16.
The cpu node process load is 9%, the memory usage is 2%.

[Qiujiao_W_Intel]

Hi david, any error message in MCU logs? BTW, currently MCU's recommend platforms are centos 7.2 and ubuntu 14.04

[david_c_1]

Hi,

The only warnings I get I have put in my first post: WARN: ClusterManager - schedule failed, purpose: portal task: 76924429298 reason: No worker available, all in full load. 

However, a workaround is to force a return true in  "isWorkerAvailable" function in the scheduler.js.

In this way I can run woogeen.

[Qiujiao_W_Intel]

Hi, please provide following information:

1. did you run all MCU modules on one device?if no, please provide your deployment, which modules run on which device, we cannot guarantee this kind of deployment work well cause currently Ubuntu 16.04 is not our support platform

2. if all modules for MCU run on Ubuntu 7.2, whether the same error still happen

3. please try run MCU on Ubuntu 14.04 not on 16.04, and check whether it worked well

4. Please provide all MCU's logs

[david_c_1]

Hi,

I confirm that all modules are running  on a single device. 

The problem occurs both on ubuntu 16.04 and CentOs 7.2. I have not tried Ubuntu 14.04 yet.

 I attach the log file of MCU.

[Qiujiao_W_Intel]

Hi, whether 1) MCU once worked normally and became abnormal at some time; or 2) MCU never well worked.

If it is case 1),  open ClusterManager module log level to DEBUG in cluster_manager/log4js_configuration.json and see whether Portal disconnected and some other unexpected conditions happened.

If it is case 2), please a) double check rabbitMQ server started correctly and TCP port 5672 is open on firewall; b) double check the rabbit.host configuration item in portal/portal.toml(and other components); 

[david_c_1]

Hi,

the woogeen service starts and runs correctly until I load the basic example in a browser's page. 

After that I get the error.

I attach the Debug log level but I do not see any significant differences compared to what is already attached in the previous post.

[Qiujiao_W_Intel]

we mean MCU totally work well in previous version like v3.3? or MCU never work well in your environment. BTW, please compress logs folder and provide whole log files

[david_c_1]

The v3.3 work well on the same device where 3.4 don't work.

I attach the full log dir.

[Lei_Z_Intel1]

Please share us the console output message of "top" command on portal node machine, we need check whether that's not expected as the performance collect script.

[david_c_1]

Hi,

Attached required top screenshot

[Lei_Z_Intel1]

Seems your "top" output a little bit different than others, it use comma instead of dot to represent decimal point. It doesn't pass our regular expression to collect CPU usage.

[david_c_1]

Hi Lei,

I hope you make a fix in version 3.4.1

[Lei_Z_Intel1]

Will fix in coming major version v3.5. Thanks for the patience.