Hello,
I have one Advanced Installer project that builds a single MSI.
The purpose of this installer is:
To deploy different application artifacts and configuration files based on the User selection on ComboBox first- Site and second ComboBox - IP (Station) during installation.
Everything is dynamic and driven by Git structure and JSON mapping files.
I would like architectural feedback on whether my current implementation is considered good practice and scalable.
________________________________________
OVERALL INSTALLATION FLOW
When the MSI runs:
1. The user selects:
o Site (from ComboBox)
o IP (from ComboBox)
2. Based on the selection:
o A custom action runs
o It queries SQL tables embedded in the MSI
o It determines which applications (artifacts) should be installed
o It sets MSI properties accordingly
Example:
If user selects:
- Site -> A
- IP -> A
Based on the SQL tables in advanced installer Custom action sets:
INSTALL_APP1 = 1
INSTALL_APP2 = 1
INSTALL_APP3= 0
0= NOT INSTALLED, 1 = INSTALL.
Then in the Feature Tab in the Advanced installer :
Install if: INSTALL_A = 1
So installation is fully controlled by properties set at runtime.
________________________________________
PART 1 — ARTIFACTS
Purpose
Deploy application artifacts based on:
• Site
• Station type
• IP
How It Works
During CI:
1. Artifacts are downloaded.
2. All artifacts are linked into the AIP in Files & Folders tab .
3. Sync folders are created to the original location in the server in Files & Folders tab.
4. One Feature per artifact is created In Organization Tab
Example:
APP1
APP2
APP3
APP4
5. SQL tables (.idt files) are generated dynamically.
The SQL tables are generated based on:
• Git structure:
Site
└── hosts
└── IP folders
• JSON mapping file:
Example:
"IOS": {
"Stations": [
"Station1:192.168.0.150",
"Station2:192.168.0.102"
],
"Artifacts": [
"APP1",
"APP2",
"APP3",
"APP4"
]
}
The JSON defines:
• Station name
• IP
• Artifacts relevant to that station type
From this:
CI generates SQL tables:
• Sites => List of sites available
• Stations => list of station available
• SiteStations => relation table sites stations
• StationsApps => which app to deploy on what station
These tables define what should be installed for each Site/IP combination.
________________________________________
Runtime Behavior
At install time:
1. User selects Site and IP from the Comboboxes.
2. Custom action queries SQL tables.
3. Custom action sets MSI properties what to install = 1 what not to install = 0:
APP1 = 1
APP2 = 0
...
4. Each artifact Feature uses property condition:
Install if APP1 = 1
So artifacts are controlled entirely by runtime SQL logic.
________________________________________
PART 2 — CONFIGURATIONS (Separate Problem)
Configurations are handled differently.
Configuration artifact structure:
ProjectName
└── IP
└── Bagira
(configuration files)
________________________________________
Current Configuration Mechanism
For configurations, during CI I generate:
One Feature per Site + IP combination.
Example:
SITE-A_192_168_0_150
SITE-A_192_168_0_151
SITE-B_192_168_0_152
...
For each Site/IP:
1. Feature is created In Organization Tab with the SITE-A_192_168_0_150 name.
2. Property is created representing default deployment location of Configurations example property : SITE-A_192_168_0_150 : “C:\IT “.
3. Link folder in Files & Folders Tab with configuration location in the git .
4. Feature condition is set:
Example:
Installed if SITE_SELECT = 1 AND IP_SELECT = 1
At runtime:
• User selects Site and IP.
• Properties are set like:
SITE_SELECT = 1
IP_SELECT = 2
• The matching Site_IP Feature installs.
• All other Site_IP Features are skipped Because they don’t answer the conditions.
________________________________________
The Scaling Issue
If one Site has 400 IPs:
→ 400 Features are generated.
This causes:
• Very large Feature tree
• Very large AIP
• CI must dynamically:
o Create Features
o Create Properties
o Add Folders
o Add Syncs
o Set Conditions
• Harder maintenance over time
________________________________________
Summary of Current Design
Artifacts Section:
• One Feature per artifact
• Runtime SQL query decides what installs
• Controlled by properties like INSTALL_A1
Configurations Section:
• One Feature per Site + IP
• Runtime selection controls which Feature installs
• Can result in hundreds of Features
________________________________________
My Concerns
1. Is using Features to represent Site/IP instances considered correct practice?
2. Is dynamically generating hundreds of Features via CI considered acceptable design?
3. Is there a more scalable architectural pattern for this kind of dynamic deployment model?
4. Am I misusing Features for runtime logic decisions?
Everything must:
• Remain dynamic
• Use built-in Advanced Installer functionality
• Better not use custom code outside MSI logic such as custom Actions
I would appreciate architectural feedback rather than specific implementation steps.
Thank you.
FOLLOW US
Get the latest news in Application Packaging