Ganesh Maharaj Mahalingam ganeshmaharaj

Resources

There are two ways you can talk to the KUBE API.

Start a kube proxy server which will act as a reverse proxy for the client.

	From a154610ce564462c5a64988b328578ac50f5e898 Mon Sep 17 00:00:00 2001
	From: Ganesh Maharaj Mahalingam <[email protected]>
	Date: Wed, 19 Sep 2018 23:15:03 -0700
	Subject: [PATCH] test glog enabling

	Signed-off-by: Ganesh Maharaj Mahalingam <[email protected]>
	---
	cluster/examples/kubernetes/ceph/operator.yaml \| 4 ++--
	cluster/examples/kubernetes/cockroachdb/operator.yaml \| 4 ++--
	cluster/examples/kubernetes/minio/operator.yaml \| 4 ++--

	The below changes are needed to get helm & tiller working with an existing kubernetes system.

	kubectl create serviceaccount --namespace kube-system tiller
	kubectl create clusterrolebinding tiller-cluster-rule --clusterrole=cluster-admin --serviceaccount=kube-system:tiller
	kubectl patch deploy --namespace kube-system tiller-deploy -p '{"spec":{"template":{"spec":{"serviceAccount":"tiller"}}}}'

	- Ceph has a script 'run-make-check.sh' https://github.com/ceph/ceph/blob/master/run-make-check.sh which is used to build & run
	a tiny ceph cluster in memory and run tests against it. Ceph CI runs this test inside a VM they spawn on a cloud environment.

	* When doing this within a traditional docker container, the test completes all the time but there are random failures each time
	leading to a false positive which doesn't help the case where i am testing a change for regressions. I suspect this is cause
	of a single kernel and there is something interesting happening when the test is being run. There is no data to backup this
	comment of mine.
	* Running it inside a traditional VM is extremely slow and becomes a no-go.
	* I tried testing this within a Clear Container to see if I can get the abstraction from the host kernel and still get it faster
	than a traditional VM. in my tests the build takes as long as it does on a VM but the test never completes. the whole container

	#!/usr/bin/env python

	"""
	Tiny program to repeateadly download a random object from a swift container
	and compare its hash with what is listed in the container
	"""

	import argparse
	from random import choice, randint
	import sys

	Problem statement: A stable release of Bifrost can break due to dependencies changing.

	Background:
	We had a bifrost VM which was responsible for deploying baremetals in our lab. We had a version of bifrost locked down
	prior to Mitaka release and it was working fine until we had to tear down the VM and re-create it. We found that
	by default ‘master’ version of ironic sources was used in the setup. This lead to the VM not being setup properly.
	On suggestions from the community, we pegged bifrost, ironic and ironic-client to ‘stable/mitaka’ and that
	worked for two weeks. On another re-build of the VM, the IPA coreos image failed to boot on the system and further
	analysis showed that the IPA image used is that of ‘master’ and not ‘stable/mitaka’.

	GENERAL
	- a reference diagram and how a general layout would look like? (I can draw out the one i have if that will help)
	- A bifrost deploy for dummies (step-by-step install guide) using the above layout. (What you want to configure, where, why, how)
	- Eg: if you use ubuntu, this is what you change
	- If you want to use your own DHCP server, this is what you change.
	- if you want to use static IPs, this is what you change/modify/add to get the machines the right data
	- Please correct me on this, if bifrost runs the DHCP server it is best to set dhcp_pool_start, end, static mask etc.. Would be nice to say that in the above guide.
	- Also i think the UUID we use in the baremetal.json is made up (random generated) and has nothing to do with the UUID of the machine you find the BIOS. Would be nice to state this as well in the document. :)
	- When an install is called, explain what happens. What the playbook does, what roles are called and what it the outcome of that.
	- various roles used, but READMEs arent li

	#! /usr/bin/python

	import argparse
	import os
	import sys
	import xml.etree.ElementTree as ET
	import xlsxwriter


	parser = argparse.ArgumentParser()