Mobile code Offloading

[asadch]

I've started working on mobile application, which can offload computationally expensive code to cloudlet and get the result back on mobile application. Please guide about the depend-abilities how to apply the task? Which cloudlet environment should be used and configure the cloudlet.Will be thankful if any one can help.

[teiszler]

Hello @asadch,

I think the best way to get started would be to look at the README for the elijah-openstack project (https://github.com/cmusatyalab/elijah-openstack). If you follow the steps there, you should have an OpenStack environment running our cloudlet extensions. From there, you can follow instructions from the elijah-provisioning README to create a Base VM image and an overlay that has the backend server component of your mobile application running on it. You can then synthesis an instance of this VM from the Cloudlet panel in the OpenStack dashboard and once you have assigned that instance a floating public IP, you should be able to have your mobile application talk to it.

I realize there is a lot of information there so I will try to break it down a little bit and if you have any other questions feel free to ask.

1. The elijah-openstack project mainly consists of a plugin for OpenStack (currently we are using the Kilo release) that introduces a Cloudlet panel and the associated functionality.
2. elijah-provisioning is a python library that encapsulates a modified QEMU binary and has additional features that make cloudlets viable such as rapid provisioning (through the use a base VM image and an overlay, only the delta between the two needs to be synthesized) and adaptive handoff between cloudlets
3. elijah-openstack will install the cloudlet library from elijah-provisioning in addition to OpenStack Kilo

Hopefully there is enough here for you to get started. Let me know if you have any other questions.

Regards,
Tom

[asadch]

Hi Tom,
Thanks for Your Response. I'm configuring the environment, will let you know once done with configuration.

Regards
Asad

[asadch]

Hi Tom,
Can you please help, how to perform this step. I'm stuck at this step.

Step 5. Configure the Ansible script.
Configure OpenStack variables

Ensure the variables defined in roles/openstack-controller/vars/main.yml and roles/openstack-compute/vars/main.yml are satisfactory.

    Specifically you should ensure that the interfaces defined by pub_iface and flat_iface are valid. pub_iface is used by the physical machine to reach the internet. flat_iface is used by OpenStack in order to communicate with the virtual machines running on the host.
    If you only have a single network interface, make sure that flat_iface is something unique as a virtual interface with that name will be created when the one_nic variable is True.
    By default, the script is setup to install all OpenStack components on a single node that has a single NIC card. If you have two network interface cards, set one_nic to False in openstack-kilo.yaml. If you are setting up a multi-node cluster where the compute node will reside on a separate host(s) from the controller, you must configure additional hosts in the hosts/inventory, change openstack-kilo.yaml to reflect which hosts should be compute nodes, and ensure that the single_node variable is set to False. You will also need to change the controller IP and hostname in the var/main.yml to reflect those of the controller.<img src="/uploads/default/original/1X/8dec0fa9f0c6578879c670f50e2a00c37dac0ff2.png" width="689" height="152">

[teiszler]

Does the physical machine you are installing on have a two network interface cards? If so, specify one to be used as the flat (internal VM communication) interface and the other for the public (i.e. access to the internet) interface. If your machine only has a single NIC, then just choose a unique name for the flat interface (something like veth1) and the Ansible script will create a dummy interface for this purpose. OpenStack prefers to have to networks, one for inter-VM communication/management and the other for public/internet connectivity.

Hope this helps,
Tom

[asadch]

Hi teiszier my machine has one NIC. I'm stuck at this point, Cursor keep blinking at point " install library using fabric"

The snapshot of main.yml is given as:

please comment.

[teiszler]

@asadch,

A few things.

1. You should change your variable flat_iface to something other than eth0. Since you only have a single NIC, this can be anything, but it should be distinct from eth0. Try 'eth1' or 'veth0'.

2. The likely reason that it is stuck on this step is because you did not properly specify the local user/password for the installation of elijah-provisioning. In 'roles/cloudlet/vars/main.yml', there are variables for a local user and password. These must correspond to an actual user and password on your machine so that the fabric installation will work. I realize this is not the ideal solution for an automated script like Ansible, but I didn't want to pull all the steps out of fabric and convert all that work into Ansible tasks. So this username/password is necessary for the installation via fabric as the script from the elijah-provisioning repo expects it to be installed by a user and not root.

Hope this help!

Thanks,
Tom

[zetelight]

Hi there, please allow me to ask a question about installation.
When I installed the Openstack++ by following the README.
At the 6th step

ansible-playbook -i ./hosts openstackpp.yaml

I encountered several issues. Please see below.

TASK [openstack-controller : (HEAT) give heat_domain_admin admin privs]("give hadmin heat_stack_owner privs" has the same output) ********
fatal: [localhost]: FAILED! => {"changed": true, "cmd": ["openstack", "--os-token", "admin", "--os-url", "http://localhost:35357/v2.0", "role", "add", "--project", "admin", "--user", "heat_domain_admin", "admin"], "delta": "0:00:00.448095", "end": "2018-05-26 11:15:06.430856", "msg": "non-zero return code", "rc": 1, "start": "2018-05-26 11:15:05.982761", "stderr": "ERROR: openstack Conflict occurred attempting to store role grant - User f9a861be3fd94631acb6abd2f5395b75 already has role 7ce0923dc21b4971a81d149059f6d4f9 in tenant 34cfc118a3af4857a3896251ea3a784d (HTTP 409) (Request-ID: req-9d834613-72b0-42f8-bcd1-89ed6be7e293)", "stderr_lines": ["ERROR: openstack Conflict occurred attempting to store role grant - User f9a861be3fd94631acb6abd2f5395b75 already has role 7ce0923dc21b4971a81d149059f6d4f9 in tenant 34cfc118a3af4857a3896251ea3a784d (HTTP 409) (Request-ID: req-9d834613-72b0-42f8-bcd1-89ed6be7e293)"], "stdout": "", "stdout_lines": []}

Finally, I also got

TASK [Gathering Facts] *********************************************************
ok: [localhost]
ERROR! Cannot reset connection:
Control socket connect(/root/.ansible/cp/346ca06965): No such file or directory

What I did:
1. At the first time, I forgot to be the root user and I use "sudo" run all commands, but then I found it and used the root user to re-install all stuff by following the README.
2. When I got this error, I did try to re-install the whole thing by using the ansible-playbook command above 2 or 3 times.
3. I checked the folder "/root/.ansible/cp/", and found it is empty.

More config info:

rabbit_pw: admin
mysql_pw: admin
db_pw: admin
keystone_admin_pw: admin
admin_token: admin
metadata_secret: admin
openstack_controller_node: localhost
controller_ip: 127.0.0.1
pub_iface: eth0
flat_iface: veth0

OS: Ubuntu 14.04 LTS (I have two system on my laptop: Ubuntu 18.04 LTS)
NIC: one.
A little extra info:
When I installed 14.04 along with 18.04, the network service didn't install well due to dual systems, so I installed an extra network driver from Intel.

Any suggestions or tips will be appreciated! Do you think I have to re-install Ubuntu 14.04 LTS and do everything again?
Thank you!

UPDATED
I just re-installed my Ubuntu 14.04LTS and did the fresh installation again.
At this time, I don't any errors except the last line

ERROR! Cannot reset connection:
Control socket connect(/root/.ansible/cp/346ca06965): No such file or directory

I am guessing the reason is that I even didn't set up any connections so it cannot be reset? If not, does it matter? Thank you!

Best,

Adam

[teiszler]

@zetelight,

This occurs in the following task of the cloudlet-gateway role:

• name: (GATEWAY) sever ssh connection and reconnect so that environment variables are loaded
  meta: reset_connection

I am trying to force Ansible to use a new ssh connection so that the environment variables that were just added to the system can be reloaded. Unfortunately due to the following Ansible bug, reset_connection isn't working properly:

Issue: reset_connection targets wrong ControlPath

If you launch the playbook again it should continue on past that step since you will be using a fresh SSH connection. You can also constrain it to that roles by using '--tags gateway' when you launch the playbook. This will skip all of other roles (openstack controller/compute, etc) and only perform the steps for the gateway role.

RE: The 409 HTTP response in the (HEAT) give heat_domain_admin admin privs task

This can be ignored. This is because you must've already run this portion of the playbook before and when you attempt to give privileges to a user that it already has, OpenStack returns a 409 and I instruct Ansible to ignore that error and carry on with the steps. This also explains why you no longer saw it on a fresh install.

Regards,
Tom