You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Aug 13, 2021. It is now read-only.
Communication Track: Louis, Oskar (Status)
Incentive Track: Fabio, Vojtech (OIVLabs)
Description
Context
Swarm has a simulation framework used to run integration tests across multiple swarm nodes. This only works on a single machine using an 'in process' adapter. This provides a fast feedback loop.
Swarm also has helm charts which can be used to deploy a swarm cluster within kubernetes. Once a cluster is deployed, we can run smoke tests to verify that the basic components are working as intended. This has a slower feedback loop than the simulation framework but is more powerful in terms of logging/metrics aggregation.
It would be good if we could have a mix of both, meaning that a developer can write the simulations locally (in process adapter), and once he/she feels comfortable with the test, we can switch to a remote adapter and run the same test against remote auto generated nodes.
Acceptance criteria
As a swarm dev, I want to run a simple simulation test against remote nodes
As a swarm dev, I want to run all existing simulation tests against remote nodes
As a swarm dev, I want to see the logs of all nodes, when using the simulation framework
As a swarm dev, I want to see all the metrics generated by swarm nodes, when using the simulation framework.
As a swarm dev, I want to see traces generated by swarm nodes, when using the simulation framework.
As a swarm dev, I want to pause simulations when an error appears so that I can debug the problem.
As a swarm dev, I want to cleanup dangling remote nodes generated by the simulation framework.
Bonus
Nice to have:
As a swarm dev, I want to simulate churn rate using the simulation framework.
As a swarm dev, I want to smulate network latency using the simulation framework.
Rationale
As a swarm developer I would like to target my tests against remote swarm nodes by using the simulation framework.
Owner
Rafael @skylenet
Stakeholder Point of Contact
Communication Track: Louis, Oskar (Status)
Incentive Track: Fabio, Vojtech (OIVLabs)
Description
Context
Swarm has a simulation framework used to run integration tests across multiple swarm nodes. This only works on a single machine using an 'in process' adapter. This provides a fast feedback loop.
Swarm also has helm charts which can be used to deploy a swarm cluster within kubernetes. Once a cluster is deployed, we can run smoke tests to verify that the basic components are working as intended. This has a slower feedback loop than the simulation framework but is more powerful in terms of logging/metrics aggregation.
It would be good if we could have a mix of both, meaning that a developer can write the simulations locally (in process adapter), and once he/she feels comfortable with the test, we can switch to a remote adapter and run the same test against remote auto generated nodes.
Acceptance criteria
Bonus
Nice to have:
Proof