Werk #12389: HW/SW Inventory: Change raw tree structure
| Component | HW/SW inventory | ||||||||
| Title | HW/SW Inventory: Change raw tree structure | ||||||||
| Date | Jul 1, 2021 | ||||||||
| Level | Trivial Change | ||||||||
| Class | New Feature | ||||||||
| Compatibility | Incompatible - Manual interaction might be required | ||||||||
| Checkmk versions & editions |
|
This only affects users which process raw trees from HW/SW Inventory system via
- files "var/check_mk/inventory/HOSTNAME" or
- export hooks, similar to the ruleset "Export List of Software packages as CSV file" or
- (Web-)API, see also werk 3585.
Other functionalities like "HW/SW Inventory History" or using single values or tables from "HW/SW Inventory" trees in views are not affected.
Old raw trees are migrated on-the-fly and the new structure is used.
Summa summarum: If a user does not access raw trees from above topics then it seems that nothing has changed.
Details: The previous raw tree structure of the "HW/SW Inventory Tree" had some disadvantages, ie. Python dicts or lists were used for two different entities:
- single values or categories resp.
- tables or indexed categories
Example: A raw tree looked like:
{
"first-path-to": {
"specific-single_values": {
"single0": "Value 0",
"single1": "Value 1",
...
},
"single0": "Value 0",
"single1": "Value 1",
...
"table": [{"col0": "Value 0", "col1": "Value 1"}, ...],
"category0": {...},
"category1": {...},
...
}
"second-path-to": [
{
"indexed0-specific-single_values": {
"single0": "Value 0",
"single1": "Value 1",
...
},
"single0": "Value 0",
"single1": "Value 1",
...
"indexed0-table": [{"col0": "Value 0", "col1": "Value 1"}, ...],
"indexed0-category0": {...},
"indexed0-category1": {...},
},
{
"indexed1-specific-single_values": {
"single0": "Value 0",
"single1": "Value 1",
...
},
"single0": "Value 0",
"single1": "Value 1",
...
"indexed1-table": [{"col0": "Value 0", "col1": "Value 1"}, ...],
"indexed1-category0": {...},
"indexed1-category1": {...},
},
...
],
}
Moreover it was not clear and inflexible to which entity a path led:
- path to single values or
- path to categories or
- path to tables or
- path to single values and categories or
- path to tables and categories or
- path to single values and tables or
- path to single values, tables and categories.
Example: It was impossible to have different entities below ["first-path-to", "table"].
The indexed categories are also cleaned up: there is no advantage of these list-based categories.
Now the raw tree structure has basically the following form:
{
"Attributes": {
"Pairs": {...},
},
"Table": {
"KeyColumns": [...],
"Rows": [...],
}
"Nodes": {
"path-to": {
"Attributes": {
"Pairs": {...},
},
"Table": {
"KeyColumns": [...],
"Rows": [...],
},
"Nodes": {
"category": {...},
...
},
},
...
},
}
The main advantage of this new tree structure is that at every level each entity is encapsulated. Therefore single values ("Attributes"), a table ("Table") or categories ("Nodes") can "live" side-by-side. Beside other improvements like easier understanding of the raw tree structure the de/serializing process is less error-prone. Also the role of paths is clearer than before.
Example: On a Linux machine with installed "mk_inventory" and 16000 entries the raw tree file increases from 347977 bytes to 349904 bytes, ie. 1927 bytes.
The file size of small raw trees strikingly increase but for medium or large raw trees the file size does not increase as much (see above example).