[Fiware-ops-dev] issues with Kurento in the Lab

[Saúl Pablo Labajo Izquierdo]

Hi, Sorry for the delay  and thank you for the detailed study of the errors happening with your setup. In it, we could recognize two issues which should be further studied as they are the most probable cause for the problems you have been facing:

The most relevant issue is the first WARN message as described in your document. This is the text of the warning:

WARN (...) : [KurentoClient]  Exception trying to reconnect to server  <ws://> ws://<IP>:<Port>/kurento. Retrying in 5000 millis

Followed by a Java exception:

org.kurento.jsonrpc.JsonRpcException: [KurentoClient]  Exception connecting to WebSocket server  <ws://> ws://<IP>:<Port>/kurento
    at (...)
Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connection refused: localhost/<IP>:<Port>
    at (...)

This happens because the connectivity with the KMS server instance has been interrupted. In other words: the Kurento Client library for Java applications, which manages all RPC between the application and KMS, has detected that KMS became unresponsive. The Client library is programmed so it will start trying to reconnect with KMS whenever this situation is encountered.

Reasons for this can vary from real network connectivity issues, to an unexpected failure in the KMS process. Best action would be to check if KMS is actually listening continuously at the expected port (in this case, 8888), and also access the logs of the Kurento Media Server (by default located at /var/log/kurento-media-server/...) and check if the KMS process is dying or complaining about some error.

Apart from this, there is another point in your study which is very telling: none of the Kurento tutorial apps should get more than a couple seconds to launch; maybe it would be acceptable that it took around 10 seconds, but 80, 100, 180 or even 200 seconds are way, way above the expected initialization times for these applications. This is assuming an already populated cache of Maven packages; of course the _very_ first initialization of the app could be much longer due to the downloading of all Maven dependencies from the Maven Central repositories, and that would depend on the overall network bandwidth and latency of the server machine.

In your screenshots there are only log messages of the "INFO" level; you can increment the logging verbosity of the application by means of the "application.properties" file:

Edit "src/main/resources/application.properties" and add these lines:

logging.level.root=INFO
logging.level.org.apache=WARN
logging.level.org.springframework=WARN
logging.level.org.kurento=DEBUG

You can also use "DEBUG" for the apache and springframework components, if you want to increase the amount of total information you get from both the Kurento Client library and the Java Spring and Apache components.

Regards



 

De: 'Enrique Sucar S.' via Fiware [mailto:fiware at naevatec.com] 
Enviado el: miércoles, 15 de noviembre de 2017 17:31
Para: Juan Angel Fuentes <jafuentes at naevatec.com>; fiware-ops-dev <fiware-ops-dev at lists.fiware.org>
CC: Federico Michele Facca <federico.facca at martel-innovate.com>; Juan Navarro <jnavarro at naevatec.com>; fiware at naevatec.com; anibalandrade <anibalandrade at inaoep.mx>; Miguel Palacios <mkpalacio at gmail.com>
Asunto: Re: issues with Kurento in the Lab

 

Hello,

The tests and results (errors) are detailed in this document:

link: https://docs.google.com/document/d/1QeSTu4-4FXixSdtnAcT9aY0H0W-4zmvwhKGSWFN9qP8/edit?usp=sharing

If you have any questions / solutions please copy Anibal and Miguel.

Thanks for your help!

Best regards,
Enrique





El 11/15/17 a las 10:20 AM, Juan Angel Fuentes escribió:

Hi,

 

No problem at all  to have it in Spanish.

 

Regards

 

--
Juan Ángel Fuentes
Telf. 644 526 530
jafuentes at naevatec.com <mailto:jafuentes at naevatec.com> 

 

On Nov 15, 2017 15:56, "Enrique Sucar S." <esucar at inaoep.mx <mailto:esucar at inaoep.mx> > wrote:

Hello Juan,

Can I send you the report in Spanish? (the student did it in Spanish, I can translate it if necessary)

Regards,

Enrique

 

 

El 11/13/17 a las 12:01 PM, Federico Michele Facca escribió:

Hi Juan Angel, 

Thanks a lot, and sorry for contacting this way, I lost a bit track who is the owner of Kurento in FIWARE,

and Fernando pointed me to your team.

 

@Enrique could please ask your team to send the email to  fiware-tech-help at lists.fiware.org <mailto:fiware-tech-help at lists.fiware.org>  ?

 

It would be good that any concrete info to replicate the issue could be provided.

I suppose, it may not be a problem to allow naevatec them to access the VMs.

 

Federico

Dr. Federico Michele Facca
Head of Martel Lab

Martel Innovate
Ueberlandstrasse 111 - 8600 Duebendorf (Switzerland)
0041 78 807 58 38 <tel:+41%2078%20807%2058%2038> 
0041 31 994 25 25 <tel:+41%2031%20994%2025%2025> 
Visit our all-new website at martel-innovate.com <http://martel-innovate.com> 
Follow us on twitter <https://twitter.com/martel_innovate?lang=en> 

 

 

On 13 Nov 2017, at 18:05, Juan Angel Fuentes <jafuentes at naevatec.com <mailto:jafuentes at naevatec.com> > wrote:

 

Hi Federico,

Could you please open a ticket in Jira?? meanwhile we will start working on this issue.

 

How long that it takes the issue to be exploit in the lab?? is it necessary to be using kurento all the time, or just leaving it up with no pipeline running happens ??

 

Please provide as much information as you could so we could replicate the issue. For e.g. a running instance in the lab so we could test on it.

 

Regards




--
Juan Ángel Fuentes Domínguez
jafuentes at naevatec.com <mailto:jafuentes at naevatec.com> 
Tlf. 644 526 530 <tel:644%2052%2065%2030> 

 

On 13 November 2017 at 17:15, Federico Michele Facca <federico.facca at martel-innovate.com <mailto:federico.facca at martel-innovate.com> > wrote:

Dear Juan and Juan Angel, 

some Mexican colleagues are experiencing issues with using Kurento on the FIWARE Lab.

 

Apparently, while locally everything works fine, after some time that Kurento is running in the Lab,

it stops providing proper socket connection. In our experience this could be as well related to MTU

which is 1400 on the Lab (while usually is 1500)

 

Is there any chance you can support them? Is there away to configure MTU?

Should they open a ticket in JIRA?

 

Cheers,

Federico

Dr. Federico Michele Facca
Head of Martel Lab

Martel Innovate
Ueberlandstrasse 111 - 8600 Duebendorf (Switzerland <https://maps.google.com/?q=Ueberlandstrasse+111+-+8600+Duebendorf+%28Switzerland&entry=gmail&source=g> )
0041 78 807 58 38 <tel:+41%2078%20807%2058%2038> 
0041 31 994 25 25 <tel:+41%2031%20994%2025%2025> 
Visit our all-new website at martel-innovate.com <http://martel-innovate.com/> 
Follow us on twitter <https://twitter.com/martel_innovate?lang=en> 

 

 

 

 

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.fiware.org/private/fiware-ops-dev/attachments/20171117/917a3680/attachment.html>