The Runners.Kubernetes config struct should allow a ServiceAccount to be set for client and master. This is to allow, e.g., PodSecurityPolicies, to be applied.

Would be good to allow tests to be run with anti affinity (potentially even affinity) based on a given hosts list with labels.

Add better "can we run something on nodes" checks (e.g., https://github.com/rook/rook/blob/master/pkg/operator/k8sutil/node.go) and also allow users to specify a list of tolerations and / or ServiceAccount to use for the Pod specifications.

An output option to SQLite and MySQL would be good, as being able to run SELECTs over the data is "faster" than, e.g., doing excel foo depending on which excel belt you have.

For MySQL it would just take db information to connect to a database and then even create the necessary table(s) for it.

Must come testers are:

• iperf - iperf3 works fine.
• smokeping
• siege and bombardment (use siege2csv for bombardment parsing)
• qperf

Optional testers to come:

• traceroute (multiple runs in one go)

• Allow parallel Ansible setup module calls (configurable)
• "Cache" Ansible host info (e.g., host ABC used as a server and client has the facts setup command run twice)
• Output of which servers were okay and which failed is broken/ doesn't look good as it is weirdly broken up.
• Ansible host info seems to take "forever" on Ancientt's side to get the response from the Ansible command.

As title says, the MySQL output should have an option to enable / disable the generation of the CREATE TABLE queries.

Add an Ansible runner.

The Ansible runner would need to take care of installing the test tool during runners.Runner.Prepare() func call.
The get hosts func would return info based on the inventory (ansible-inventory command can probably be used).

As of now the Data is shown bit_per_second only.
It would be really helpful if there is an additional column with "gbits_per_second".

Add siege as a tester tool.

Old ticket info:

siege and bombardment (use siege2csv for bombardment parsing)

The test results should be made available to Prometheus in some to be defined way.

TODO:

• Discuss potential implementation ways for this.
• Future question: How can ancientt be changed that it can run in a daemon / exporter like mode / way to expose Prometheus metrics?

Add Smokeping as a tester tool.

The IPerf3 Client result has an "end" result which would be good to have also as an available result / output.

Create a simple Ansible playbook to install the current dependencies prgep, iperf3 and in the future others. Differentiating between testers used would be cool.

The playbook should "compatible" with at least Deb and RHEL based OSes (e.g., use https://docs.ansible.com/ansible/latest/modules/package_module.html).

Add qperf as a testing tool.

There need to be some tests which simulate runs of the application automatically and ensure everything worked fine.

Add traceroute as a network testing tool.

Needs to be so that one error, unless "fatal" doesn't cause the whole program to hang in the end, especially when the user gives the ignoreErrors run option for a task.

This currently causes unwanted test abortions.

In that round of code change(s), it would be good to have a Status column as a "Standard" added to the results per server_host -> client_host task, e.g., wasn't able to start the server Pod will have a Status Error or so.
Problem with implementing this would mean a even higher size of the results file (especially for, e.g., CSV, Excel and others..).

Pkg: github.com/wcharczuk/go-chart

Allow creating charts for various variation "types" (e.g., per server, summary, per data column) through this output.

Following things need improvement:

• Multi arch builds
• Tar releases together with LICENSE, NOTICE and example config file
• (Mostly) automatic changelog generation

Pkg to use: https://github.com/360EntSecGroup-Skylar/excelize

We need an additional output model in form of excel sheet(.xlsx format?)

Create unit tests for each outputs, parsers, runners and testers.

The Kubernetes runner should allow to use Service IPs for the Servers instead of just the Pod IPs.

Interesting would also to allow use of hostNetwork in that change.

Currently all outputs are in the outputs/ directory of the repository.
To put it bluntly "it is getting crowded in there" with all constant and variable names.

Would be good to look into splitting it into sub pkg per output or so.

Edit: This will probably be needed for parsers, runners and testers as well..

Use, e.g., logrus for logging of when a Pod is created and so on.

It is tedious to have all those if config.ABC == nil and so on in each outputs, parsers, runners and testers.

Would be good to improve that and and add validation in the same change.

It would be worth checking out Calver to see if it makes sense to use it for the project instead of Semver.