jrafanie

• build and system directories extracted to manageiq-appliance repo.
• Make that work by building off a fork of manageiq with build and system removed
• Split manageiq-appliance(build/system) to manageiq-appliance-build and manageiq-appliance, removing the leading build/system directories.
  • Apply manageiq build directory commits to manageiq-appliance-build.
  • Apply manageiq-appliance build directory commits to manageiq-appliance-build.
  • Save off the system commits from manageiq-appliance. (NONE)
  • Delete manageiq-appliance, recreate it.
  • Apply manageiq system directory commits to manageiq-appliance.
  • Apply old manageiq-appliance system directory commits to manageiq-appliance. (NONE)
• Make the build work(like step 2) but with the 3 repos.

jrafanie

# This gem with my changes can be found here: https://github.com/ManageIQ/manageiq-decorators/compare/master...jrafanie:manageiq-decorators:zeitwerk

joerafaniello@Joes-MacBook-Pro-2 manageiq-decorators % tree lib
lib
├── manageiq
│   ├── decorators
│   │   ├── engine.rb
│   │   └── version.rb
│   └── decorators.rb
├── manageiq-decorators.rb

jrafanie

1. Determine your data directory:

psql -U root vmdb_development -c "show data_directory;"

2. Edit: DATA_DIRECTORY_FROM_STEP1/pg_hba.conf

Ensure this line is added:

local replication all trust

jrafanie

Summary:

One would expect the before and after checkout callback would wrap checkout or the undlying code that creates a new adapter and assigns it to a pool or assigns an existing one from the pool. The before and after checkout callback actually wraps verify.

Note, the behavior is the same in 5.1.x, 5.2.x, 6.0.x, 6.1.x but the code snippets come from: https://github.com/rails/rails/blob/v6.0.4/activerecord/lib/active_record/connection_adapters/abstract/connection_pool.rb

Therefore, there is no way to instrument diagnostic code BEFORE the actual checkout occurs.

jrafanie

On my fyre infra node, I noticed there was already an nfs PV for the registry storage and I couldn’t figure out how to export a new nfs properly so I just copied it.

[root@jrafanie-inf data]# cat ~/image-pv.yaml
apiVersion: v1
kind: PersistentVolume
metadata:
  name: registry-storage
spec:
  capacity:
    storage: 200Gi

jrafanie

* Install go

brew install go@1.13
echo 'export PATH="/usr/local/opt/go@1.13/bin:$PATH"' >> ~/.zshrc
export PATH="/usr/local/opt/go@1.13/bin:$PATH"
go version

* Install the operator-sdk:
(Note, redhat laptop is using operator-sdk 0.19.3, go, 1.13 and bin/build works but the operator-sdk generate commands were not tested)

jrafanie

irb(main):021:0> MiqWorker.descendants.each {|w| puts "Class: #{w.name.ljust(80)} Deployment: #{w.new.worker_deployment_name.ljust(50)} Advanced settings: #{w.path_to_my_worker_settings}"}; nil

Class: MiqUiWorker                                                                      Deployment: 1-ui                                               Advanced settings: [:worker_base, :ui_worker]
Class: MiqQueueWorkerBase                                                               Deployment: 1-queue-worker-base                                Advanced settings: [:worker_base, :queue_worker_base]
Class: MiqCockpitWsWorker                                                               Deployment: 1-cockpit-ws                                       Advanced settings: [:worker_base, :cockpit_ws_worker]
Class: MiqRemoteConsoleWorker                                                           Deployment: 1-remote-console                                   Advanced settings: [:worker_base, :remote_console_worker]
Class: MiqSch

jrafanie

How to setup replication in pods with postgresql 12 (very similar to 10):

1) Do the following steps for each region: example: 99 for global, 1 for region
In your local manageiq-pods repo, modify the deploy/crds/manageiq.org_v1alpha1_manageiq_cr.yaml with this information:

...
spec:
  applicationDomain: <unique name>
  postgresqlImageName: <user/repository_of_pg_12_image such as 'jrafanie/postgresql'>
  postgresqlImageTag: <tag such as 'latest'>

jrafanie

Thread.abort_on_exception = true
STDOUT.sync = true
STDERR.sync = true

module DeploymentCheck
  def watch_pod_options
    {
    :label_selector   => "app=manageiq",
    :namespace        => my_namespace,
    # :label_selector => "app=manageiq,worker=true"

jrafanie

require_relative 'config/environment'

Notification.includes(:subject).all.each do |n|
  begin
    # Options could have no longer available activerecord/activemodel classes from rails 5.0.0 -> 5.0.6, which will raise an ArgumentError on deserialization
    # Fixing these to use 5.0.7+ smaller serialization format without all the private attributes is too hard, especailly since you can only read their YAML strings
    # and make changes as a string, not as a Hash.
    opts = n.options
  rescue ArgumentError
    new_value = n.subject.try(:name) || n.subject.try(:description)