crash-diagnostics

crash-diagnostics copied to clipboard

### What is happening? While running the tests, the created `kind` clusters get added to the default location of the kube config `$HOME/.kube/config` and the current context set in the...

srm09

SSH is not always available or ideal, in many cases using a DaemonSet or Jobs to collect logs is more practical e.g. https://github.com/kubernetes-sigs/cluster-api/issues/3344

moshloop

When Crashd captures files and saves them in the working directory, it currently overwrites existing files or keep adding files to the same working directory from previous run. This can...

vladimirvivien

Currently when the SSH container is started for tests, it uses the same mounted volume location for all tests. This can cause some tests to fail especially when ran in...

vladimirvivien

As a script writer, I should be able to specify connectivity/retry parameters for kube-related commands. For instance, `kube_config` should be updated to support max-retries and timeout as follow: ```python kube_config(path="./kube/config",...

vladimirvivien

Currently the crash-diagnostics codebase uses [logrus](https://github.com/Sirupsen/logrus) which has some potential issues with goroutines. Upstream CAPI uses `logr` which is an interface based logging mechanism. Possibly switch to use `logr`

srm09

As a script developer, I should be able to specify required program used in the script. If the program is not found or not installed, the script should fail fast....

vladimirvivien

srm09

Currently, the code ignores and does not validate host keys during an SSH/SCP operation. While this allows Crashd scripts to run quietly, it can be viewed as a security issue...

vladimirvivien

etcd crash logs

1

There are two really useful text snpipets we can collect about etcd over ssh # Note these have to happen on CAPI Masters ## 1) results of etcd perf, making...

jayunit100