Simulate the accidental deletion of a nat gateway by a human operator. This can also be used to simulate isolation scenarios.

Need to be able to:

• get nat gateways with a given vcn id
• filter the nat gateway using given a set of metadata
• delete a nat gateway using a given nat gateway id
• rollback the deletion of the nat gateway

1. Update and clean up README
2. Add folder for documentation
   • Python setup
   • Add user setup
   • Add developer setup
3. Update requirements.txt (chaostoolkit-lib version), requirements-dev.txt (pytest version)

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Description

The chaostoolkit-oci project sets some coding standards e.g. pep8 and also requires packages. These are documented a little bit all over the place either in the chaostool-oci, the chaostoolkit itself or sometimes we have to check the Python documentation.

It would be good if these could be documented in the project here to facilitate developer set up, adoption of common standards etc to facilitate the rapid integration of new contributors.

Potential Configuration

# Copy-paste your configuration here. 

References

A useful experiment is to product a machine network isolation on a compute resource.

The reason for the experiment would be not only to see how a cluster would behave if it looses a node (that can be done by stopping the resource) but to see how a resource would behave if it looses its cluster.

Hi,
I am interested in using and contributing to this project. Would you be able to clarify who owns the copyright? I don't see any notices in the repo.
Thanks a lot for your help,
Mark Nelson

The capability of stopping an instance_pool, analogous to AWS asg.

This will allow to experiment with shutting down (stopping) several compute resources at once from a given group.

This may not be possible but it probably is, to create a generic probe for counting initially, the block volumes, block boot volumes, backups, etc.

Although it would be just one probe, it's possible to create a decorator to use whatever sdk function is needed.

Simulate an accidental routing table deletion.

Simulate the accidental deletion of a routing table by a human operator. This can also be used to simulate isolation scenarios.

Need to be able to:

1. get route tables with a given vcn id
2. filter the route table using given a set of metadata
3. delete a route table a given route table id
4. delete a route table using a given set of metadata
5. rollback the deletion of the route table

Oracle cloud has three concepts, Region, AD (Availability Domain) and FD (failure domain), resources in one FD are independent from resources in a different FD, if one fails the rest remains fine, hence the name.

This action needs to be able to stop/force stop, an entire set of resources in a given FD.

In order to be more useful, I would add a probe that counts the number of FDs in a given AD, this is in the hypothetical case we have resources in two failure domains, and we want to shutdown one of them as part of the experiment or not shutdown anything if there is one FD, implying we don't have any failover scenario or HA.

In cases where instances do not have a public ip address, they use a nat gateway to access the internet; by blocking the traffic of the nat gateway we can simulate an entire subnet failure and get insights on what would happen if we suddenly lost an entire group of nodes.

We can test:

• Monitoring system and see alerting.
• How external monitoring would alert us if the monitoring system inside the network is not able to reach the internet (if part of the experiment).

This can be an interesting experiment and it is simple to do since the nat gateway in OCI blocks traffic directly (and rollbacks if needed) without any major change in configuration.

Currently, under chaosoci, we have the following:

compute
init.py
types.py

We need to refactor and reorganize so it follows the structure of the oci-python-sdk e.g.:

chaosoci
|core
| block
|_compute
|_computemanagement
|_networking
....

Likewise, the tests need to follow the same structure.

This will make it easier to add the different probes and actions in the future.

Hey all,

Would you want me to add this extension to https://docs.chaostoolkit.org/drivers/overview/ ?