crash-diagnostics
crash-diagnostics copied to clipboard
Crash-Diagnostics (Crashd) is a tool to help investigate, analyze, and troubleshoot unresponsive or crashed Kubernetes clusters.
Besides the scope filter (see #55), a user should be able to specify node labels that filters which node the command should execute on: ``` FROM nodes:"all" COPY /var/logs labels:"kubernetes.io/hostname=control-plane"...
Currently, crash-diagnostics uses a single execution runtime that assumes all commands are executed remotely using ssh or scp. While this assumption is easy to implement, it also makes the tool...
By default, all commands listed in a crash-diag is executed on each machine specified in the FROM list. This issue is to introduce command scope to specify which machine gets...
Crashd supports `AUTHCONFIG` and `KUBECONFIG` that are used to configure remote host connections and connection to the API server respectively. Right now, each Diagnostics file can only use one connection....
Commands that execute on remote machines will retry uses SSH/SCP by default. However, these commands will retry for a set amount of time or a set number of time. This...
The script parser and its executor needs to be exposed as clear programmable Go API, via their respective packages, to allow developers to run script programmatically. This may require some...
Hi, Is there a plan to support some alternative for ssh to execute commands on Nodes? For example, I believe the Kind Node image doesn't come with an SSH server...
It would be helpful in the documentation to provide a fully featured example for `kubeadm` bootstrapped clusters. This would be a stable target as well as allow one to get...
A few ideas: * enable the user to provide the Node names; crash-diagnostics can connect to the K8s apiserver to retrieve the IPs * enable the user to provide host...