mirko-laruina / gnocchi_sys_stats

Project for the course in Cloud Computing @ UniPi

gnocchi_sys_stats

Project for the course in Cloud Computing @ UniPi

It provides an application to visualize system statistics (CPU and memory usage), stored using Gnocchi with OpenStack.

Usage

Getting a token

Connect to a node from which you can connect to OpenStack, login using the admin.sh script and request the issue of a new token.

$ source admin.sh
$ openstack token issue

Then copy the content of the id row, it will look like the following (I have wrapped the token over multiple lines for sake of visualization but usually it is not broken):

+------------+-----------------------------------------------------------------+
| Field      | Value                                                           |
+------------+-----------------------------------------------------------------+
| expires    | 2020-07-23T17:49:57+0000                                        |
| id         | gAAAAABfGb-1D4yI2O06fJj5rJwANTAWfGyIWHLRJezjlsYQXbrOkefOtdfmfkoe|
|            | jc4mgDquIJINY4pDyZJA5q3hhkCZckZDYMu5rQWsjt2dB85YfUQVCy04cqkb-bNC|
|            | UuH-lNYlxKtnjBepZyYgaSpTItZnxHBlFyX4xKVmK8BP_apvs7EM_VQ         |
| project_id | 56db67ef8e1d4ed78c7589bdd2fd73fc                                |
| user_id    | 5baa4ea88e9c4fc192b71bbcd24fd2a3                                |
+------------+-----------------------------------------------------------------+

Running the producer

On the machine you want to monitor run the following:

$ cd src
$ TOKEN=<token>
$ python producer.py -t $TOKEN

You can also pass the following arguments to the script in case you need:

• -i: changes update interval (in seconds)
• -u: sets Gnocchi endpoint URL (defaults to the one we used in our experiments)
• --project_id: sets project id (defaults to the one we used in our experiments)
• --user_id: sets user id (defaults to the one we used in our experiments)

Running the consumer

There are three modes of operation: list, plot and dump.

To all commands shown below, you can pass the following additional arguments:

• -u: sets Gnocchi endpoint URL (defaults to the one we used in our experiments)
• --user_id: sets user id (defaults to the one we used in our experiments)
• --project_id: sets project id (defaults to the one we used in our experiments)

Use list to list all available resources along with available metrics:

$ cd src
$ TOKEN=<token>
$ python consumer.py list -t $TOKEN

Once you find the resource you want to monitor, you can copy its UUID and use the plot command to monitor one of its metrics:

$ cd src
$ TOKEN=<token>
$ python consumer.py plot <uuid> -t $TOKEN -m <metric>

You can add multiple resources by just adding their UUID after the one shown above.

You can also pass the following arguments to the script to change its behaviour:

• -g: change granularity (default second)
• -r: resample the measures to a more coarse representation (e.g. 10s, 2m, ...)
• -a: choose how to aggregate resampled measures (e.g. mean, max, min, ...)

If you prefer a command line representation of the incoming data, you can use the dump command:

$ cd src
$ TOKEN=<token>
$ python consumer.py dump <uuid> -t $TOKEN -m <metric>

NB: this command can monitor only one resource at a time

You can also pass the following arguments to the script to change its behaviour:

• -g: change granularity (default second)
• -r: resample the measures to a more coarse representation (e.g. 10s, 2m, ...)
• -a: choose how to aggregate resampled measures (e.g. mean, max, min, ...)

Bonus: running the consumer on your laptop through tunnelling

If the Gnocchi endpoint is not directly reachable from your computer's network, you can open a SSH tunnel as follows:

$ ssh -L localhost:8041:252.3.47.9:8041 [email protected]

This command connects to the user ubuntu at the (reachable) machine 172.16.3.47 and opens a local tunnel that makes all connections to localhost:8041 being redirected to 252.3.47.9:8041 (which is our Gnocchi endpoint). Remember not to close this ssh session otherwise the tunnel will be tear down.

Then you can just add -u http://localhost:8041 to the commands shown earlier to connect using the tunnel.