Here's a partial configuration to give more contex...
# gooddata-platformp
Philippe Hébert
04/05/2022, 4:33 PMHere's a partial configuration to give more context:
✅ 1
m
Michal Hauzírek
04/05/2022, 4:35 PMIs it possible that you do not have any clients/workspaces in your segment?
p
Philippe Hébert
04/05/2022, 5:00 PMI don't have client workspaces in my segment yep. Is it a requirement?
m
Michal Hauzírek
04/05/2022, 5:10 PMwell the "load to segment" action loads data to all client workspaces. Not to the master. The segment master workspace is meant more like a template from which the client workspaces are derived.
For development you usually use another workspace which we call "development" to which you load the data separatelly.
You could probably do the same thing alsi with the master workspace even though it is not thevstandard approach.
Anyways to do it you would deploy new Automated Data Distribution process and during the deployment select "load this workspace". And on the right side you can define the client ID you will use for loading this particular workspace.
p
Philippe Hébert
04/05/2022, 6:26 PMThanks for your answer Michal.
If I understood properly the advised structure is the following:
== dev ==
1. Have a development master workspace associated to a segment
2. Assign a data loading process that loads the data with a specific client identifier to the development master workspace directly
3. Release the development master workspace to a test master workspace
== test ==
1. Have a test master workspace associated to a segment
2. Run the provisioning brick
3. Assign a data loading process that loads the data to the created client workspaces
4. Release the test master workspace to production to a prod master workspace
== prod ==
(Same structure as test)
Philippe Hébert
04/05/2022, 6:28 PMa. Can you confirm I understood properly?
b. Why isn't it advised that the dev master workspace have both a workspace data loading process and a segment data loading process?
m
Michal Hauzírek
04/06/2022, 9:49 AMYes, @Philippe Hébert, with just one slight difference:
on DEV the “development master workspace” does not actually needs to be in the segment at all. It does not need to be “master workspace” of any segment. It can very well be a standalone workspace with no clients.
Michal Hauzírek
04/06/2022, 9:56 AMThe term “master” might be a but confusing, byt there are typically two kinds of workspaces:
“*development (*master) *workspace*”
• the one where you are doing the development
• the one which has specific dataload deployed (with some client_id specified)
• the one to which the release brick points (development_pid parameter)
• typically is not part of any segment
• typically is not master of any segment
“*segment master workspace*”
• the one which serves as a current template for new client workspaces in a segment
• it is created by the release brick (from the “development workspace”)
• typically does not have any data or data loads in it
• is set as a master of its segment
and then there is a third one:
“*service workspace*”
• does not have any data model and metrics in it
• does not have any data
• serves as an envelope for schedules and processes
◦ the data loading ones for load of the whole segment (ADD)
◦ the bricks for LCM - release, rollout, workspace provisioning
◦ potentially other bricks (i.e. for provisioning users etc.)
p
Philippe Hébert
04/08/2022, 10:37 PMHey Michal!
Sorry for the late answer. That's a really great answer right there!
That's the clearest description of master workspaces I have seen so far, I would really suggest to put this or a variation of it in your documentation about workspaces! 🙂