Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] Tests are not scheduling on pscheduler

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] Tests are not scheduling on pscheduler


Chronological Thread 
  • From: Andrew Lake <>
  • To: Ivan Garnizov <>, Murilo Vetter <>
  • Cc: perfsonar-user <>, Mark Feit <>
  • Subject: Re: [perfsonar-user] Tests are not scheduling on pscheduler
  • Date: Wed, 3 Apr 2019 17:10:55 +0200

Hi Murilio,

Let me try to summarize the steps you have run and what should happen so you can correct me if I missed anything and then we can delve into how to figure out what is going wrong:

1. You bring up a new perfsolnar-testpoint container
2. Inside the testpoint container you run “psconfig remote add --configure-archives http://monipe-central.rnp.br:8000/pub/config/matriz
3. Within 60 seconds the psconfig agent should pickup the new URL and start looking for tasks (i.e. no service restart needed)
4. It starts creating tasks.

It looks like at either steps 3 or 4 things are tripping-up. The most common cause for the type of behavior you are seeing is that the container has an IP address that does not map to anything in your remote JSON file. Specifically, if you were to do a DNS lookup any of the hostnames in the “addresses” section of your remote JSON, does it map to an IP address you see when you run “ifconfig” inside the container? If the answer is no, that is likely the issue and you need some additional configuration (that I can help you with if this is indeed the case).

If the hostnames do map to an IP address inside the container, then there are a couple things you can check. For one, it looks like you are using the pSConfig Web administrator, which still defaults to the old MeshConfig JSON format for backward compatibility reasons. You can get the newer format directly by running the following:

psconfig remote delete http://monipe-central.rnp.br:8000/pub/config/matriz
psconfig remote add --configure-archives "http://monipe-central.rnp.br:8000/pub/config/matriz?format=psconfig"

This might fix things if you are running into some type of translation error. If not, it will take a translation error out of the equation at least. 

If that still does not fix things you can turn on debugging in the logs by replacing INFO with DEBUG in /etc/perfsonar/psconfig/pscheduler-agent-logger.conf. We can go from there. In particular I’d be interested if we see any messages in the logs indicating it even tried to download the JSON file if it it turns out none of the above helps. 

Hope that helps get you further along.

Thanks,
Andy






On April 3, 2019 at 8:59:35 AM, Murilo Vetter () wrote:

 the system operates without the knowledge of the mesh configuration, which is added during the operation of the pS container….

 

In your case, I would suggest to prepare the psconfig configuration for  the meshes with the building of the container. Thus when started with the container psconfig will automatically parse it and configure pscheduler.

 

There might be other mean to restart a service without the use of the ‘systemctl’ command. Of course I can’t recommend that, but I guess in your case, there is nothing to loose.




Archive powered by MHonArc 2.6.19.

Top of Page