BonFIRE logo and link to main BonFIRE site

Table Of Contents

Previous topic

Troubleshooting

This Page

FAQ

Are the clocks of the various compute resources synchronized? How?

The short answer is yes. We take reasonable measures to ensure that clocks of compute resources among any testbed are within 0.5 seconds (500 milliseconds) of a reference time obtained through the Network Time Protocol (NTP). Therefore you can assume that all your VMs are synchronized at plus/minus 0.5 seconds of the NTP server BonFIRE is using. BonFIRE central services are also synchronized using the same NTP server, which ensures that the whole BonFIRE infrastructure is in sync.

Can I describe my whole experiment deployment in a single document rather than creating it in the portal or via multiple OCCI calls?

Yes, this is possible with Experiment Descriptors, using the Experiment Manager in BonFIRE. It is also possible by creating Restfully scripts if you are familiar with Ruby.

How do I control my experiment once it is running?

This depends what you mean by control. Some particular features in BonFIRE include the support for elasticity rules and dynamically changing the QoS of emulated networks on the Virtual Wall. Controlling when programs start and stop on the VMs in your experiment(s) is something that must explicitly be coded by you. For example, Restfully supports this in that it allows you to SSH into your VMs and run commands.

How can I pass on configuration parameters to VMs at deploy time?

This can be done via the contextualisation mechanism that is available via the OCCI interface used in BonFIRE. You can add any key-value paris of configuration parameters you wish, as well as specifying any custom monitoring metrics, elasticity trigger rules, additional SSH keys and post-install scripts that can be run when the VM has been deployed. For more information about how to do this, see the Contextualisation page.

Can I get my experiment VMs execute scripts automatically when they have deployed?

Yes, we refer to this as post-install, which can be set up via the contextualisation of the VM when you deploy it.

How can I get my VM to wait until contextualisation is finished before continuing with initialisation?

On ONE sites, check for the absence of /mnt/init.sh. Nn other sites, test for the presence of /tmp/contextualisation_script_done. In bash, can be tested with

[ ! -f /mnt/init.sh ] || [ -f /tmp/contextualisation_script_done ]

For more information, see the contextualisation page.

How can I create/save new VM images?

Firstly, you will have to use one of the base images in BonFIRE - you cannot create your own VM images on your local machine and load into BonFIRE. If you do require a particular image/OS that is not currently available in BonFIRE, please contact us and make a request.

When you deploy one of the base VMs in BonFIRE and SSH into it, you can install software and configure it as you like. You can save a new VM image based on this, so you don’t have to repeat this process. This is done by issuing a SAVE_AS command and then shutting down the VM. For more information about how to do this, please see the Configuring Software and Saving VMs page.

Please note that VM images you create on one testbed cannot be used on another testbed. Therefore, if you are doing a federated experiment in BonFIRE, you may need to save a VM image for each testbed you will use.

How can I reserve resources for running my experiment on?

Advanced reservation of resources in BonFIRE is not possible for the permanent resources, but can be done for the on-request resources in BonFIRE. Inria offers an online portal for doing this automatically, whilst the other testbeds need to do this manually. Please see the On Request Resources page for more details on this process.

Where can I find details of physical hosts and current usage?

General details of the infrastructures can be found here and more specific testbed details here.

There is some statistics of usage for sites, which you can read more about here.

How can I deploy a VM to a specific physical host?

The BonFIRE OCCI interface allows you to specify the host, but this is only possible on the following Open Nebula testbeds: EPCC, HLRS and Inria. For more information about how to specify the host, see for example the following OCCI tutorial.

An overview of the physical hosts available at the different testbeds is provided on the Testbed Specificities page.

How do I see which physical host my VM is running on?

If you didn’t specify the physical host to deploy your VM on, or you simply want to validate that it deployed where you specified, you can see this information in the OCCI XML you receive when deploying the resource. If you don’t use OCCI directly via cURL (for example), you can still see this information by clicking on your experiment VMs on the BonFIRE Portal.

How can I deploy multiple instances of the same VM?

The BonFIRE OCCI interface currently only supports operations for single compute resources, i.e., OCCI does not support calls to deploy X number of a particular VM. However, this functionality is supported in client tools such as the Experiment Descriptors or Restfully scripts.

Why are some basic Linux tools (e.g., less) missing from my VMs OS?

We have tried to provide as lightweight default operating system images as possible. You can easily install more applications yourself and then save the image for further use. The apt-get tool can be used to install software. For example:

apt-get update
apt-get install less

After doing these for the applications or tools you wish to install save your image and use this as your OS image for subsequent work.

How can I get monitoring data for my experiments?

First, you will need to have enabled monitoring for your experiment when you created it, which involves deploying a Zabbix Aggregator and passing on it’s IP address to the Zabbix Agents in the VMs you deploy in your experiment. For more information about how to set up the monitoring, please see the monitoring how-to page.

Asssuming you have set up monitoring for your experiments, you can always view the monitoring information via the Zabbix UI - you will find a link to this on the BonFIRE portal in the experiment view. You can download the monitoring information programmatically or exported to CSV. To be able to get the monitoring data after your experiment has finished (resources, including monitoring resources will have ben destroyed), please see this page.

Can I dynamically scale the CPU and memory assigned to my VMs?

No, BonFIRE does currently not support vertical elasticity.

How much space do storage resources occupy?

Currently all storage resources (including OS type images) occupy their full size, irrespective of the amount of data stored in them.

If the storage is created within the experiment: is it created at the target hosting node directly?

OpenNebula creates the storage on the front-end and then transfers to the worker node.

Is qcow2 supported on BonFIRE?

No, the OpenNebula version used in BonFIRE does not support qcow2.

How do I create a network with controlled bandwidth, latency etc?

There are two ways in which you can control the bandwidth in your experiment:

1: Make use of emulated networking at the Virtual Wall.

2: Make use of AutoBAHN, but note that this is only available between the EPCC and PSNC testbeds.