andrewh1978 / px-deploy Goto Github PK

When deploying in a different region, the default instance type is missing causing errors. Can a check be added to make sure the instance type is available with a more graceful failure if it is missing. Alternatively we could add a default instance for each region.

extend demo to include storage pool expand

If you run a new deployment then quickly try to SSH to the master using pxctl connect the command exists without any error output.

Sort AutoPilot events by date

If a deployment fails, clean up the artifacts and delete associated files

Ideal time for volume to hit 50% capacity is 4 mins, which allows AP enough time to react before the pvc hits 100%. Current timing is around 6 mins to 50%.

The current implentation of PX-Deploy doesn't allow for the following settings to be changes in vSphere:

1. Network to place VMs - deploys VMs on same network as template
2. VM Folder Placement - Deploys VMs in same folder as template
3. Requires a template already exist in the VMware environment that meets PX-Deploy requirements - Centos7-with-CloudInit

investigate disabled cockroachdb compression to allow for more linear growth

Check if a deployment that px-deploy thinks exists still exists; if not, clean up the associated files

https://github.com/andrewh1978/px-deploy/blob/vsphere-rp/vagrant/all-common#L63

such as My Awesome Users/Myself

px-deploy uses the AWS and GCP CLI to manage infrastructure; we should move to the SDKs

Having the DisasterRecovery option in the clusterpair template causes an error on the latest Portworx version. This needs to be removed.

I was setting up async-dr template and since I forgot to update my license in the ~/.px-deploy/defaults.yaml and not sure if there's some improvement in handling, as the create process got stuck (no timeout?) - when checking the first master's /var/log/px-deploy/ directory, i see licenses file was continually filling with:

Unable to use a TTY - input is not a terminal or the right kind of file
ERROR: Could not activate license: Error processing license: Please reduce number of nodes from 3 to 0, or add a new license
command terminated with exit code 1

I am wondering if there could be some better handling of detecting this is failing? There were no errors passed through to the user, and so after half an hour of no progress I started to wonder if something was wrong.. Asked in the #px-deploy slack channel and got a quick answer, but this could probably be handled better from a tool standpoint?

Integrate the GUI so it auto loads the cockroachdb UI

right now I think we dont do this by default which is fine, but i dont think there is a knob?

VM Provisioning takes a while, it would be nice to see progress messages, or steps being executed during the process.

Will fix to work with any k8s and any cloud.

In the case of deploying a Metro environment without setting the env var for licenses a warning should be given. If the license remains as XXXX-XXXX etc this should be caught before the deployment runs.

Somes a node fails to provision for transient reasons; we should retry, automatically or otherwise

In order to demo proxy volumes we need an NFS server running inside a cluster. I believe it should be included as standard in all Portworx deployments so it should probably be added to the k8s-master provisioning script.

templates/async-metro-backup-demo.yml does not seem to be usable?

It does not seem to install PX. https://github.com/andrewh1978/px-deploy/blob/master/templates/async-metro-backup-demo.yml#L3

I wonder if this is just misssing install-px and licenses scripts

[root@master-2 ~]# cat /var/log/px-deploy/clusterpair-dr 
No resources found in kube-system namespace.
waiting for portworx
No resources found in kube-system namespace.
waiting for portworx

Add an option to provision an environment locally on a laptop / home server using VirtualBox

assign to me and i can update.

Master node type is the same size as worker nodes.

Currently its hard coded to use "VM Network"

For example:

demo/px-central
comp/storageos

Numerous customers are interested in Oracle on Portworx. It would be good to be able to deploy in Oracle Cloud to allow for multi-cloud demos and labs.

It's currently used by the MOD, various other UK public sector and Edu oppos.

We currently have three ways to deploy central and backup - backup.yml, px-central.yml, px-backup.yml. We believe there should be two templates.

There is a use case for standalone backup and central with backup (which you may not necessarily demo). Our plan is the following:

1. Wait for the release of Central 1.0.4 which should include helm install with an option to include backup
2. Update the backup.yml template and scripts to use this new method, replicating the existing environment with latest versions of the software.
3. Deprecate current px-central.yaml (this only deploys central) and replace with the refactored backup.yml which deploys both central and backup already. Call this px-central.yml
4. Automate creation of demo clusters and the inclusion of them in the relevant dashboard in both px-backup.yml and the new px-central.yml
5. Automate adding an S3 backup location in both templates (environment var already exists and can be set in defaults.yaml).

• add kubectl patch to metro
• move storkctl install to px-install (and any other scripts that install px, eg cp4d)
• move aws stuff out of async-dr
• move aws stuff out of metro
• separate licence script
• dont apply async dr to remote clusters

right now we always deploy a new VPC. It would be good to point px-deploy at an existing VPC.