Werk #5605: HW/SW Inventory: Status data can be integrated into HW/SW Inventory tree

The HW/SW Inventory system collects static information about hardware and software of hosts which is displayed in the GUI as a tree below 'Inventory' of each host which has installed the mk_inventory plugin and enabled the rule 'Do hardware/software inventory'. If any data changes the HW/SW Inventory system detects that and calculates the differences which can be viewed below 'Inventory history'.

Now status data like free space of tablespace can be intergated into the HW/SW Inventory tree. These kind of data may change after every check cycle thus we do not calculate any differences as for the static inventory data in order to avoid filling up inventory archive and therefore slowing down the monitoring system.

We solve above problem by separating these two kind of data within the inventory plugins. - There's one tree which is filled up with inventory data and will be saved to var/check_mk/inventory as before. - A second tree for the status data will be created and saved to tmp/check_mk/structured_data

Later in the web GUI both tree objects are merged and visible below 'Inventory'. Inventory history is only calculated for the inventory data, not for the status data.

If the rule 'Do hardware/software inventory' is enabled for a host then doing status data inventory is also enabled. In this case Check_MK does a HW/SW inventory for this host after every check cycle if and only if there's a inventory plugin which fills up status data tree. Within above ruleset 'do status data inventory' can be disabled.

Notes for inventory plugin development: Previously an inventory plugin used the functions inv_tree_list and inv_tree for filling up the inventory tree, eg.

def inv_oracle_tablespaces(info):
    node = inf_tree_list("software.applications.oracle.tablespaces:")

This behaviour will be still supported, but it's recommended to use the new API, escpecially for new inventory plugins.

In order to use this new feature the inventory plugin takes the optional arguments 'inventory_tree' or 'status_data_tree', eg.

def inv_oracle_tablespaces(info, inventory_tree, status_data_tree):

There are two methods for these tree objects 'get_dict' and 'get_list' which take a path as argument. This path defines where the data should be stored, eg.

    path = "software.applications.oracle.tablespaces:"
    inv_node = inventory_tree.get_list(path)
    status_node = status_data_tree.get_list(path)

Especially if lists are filled up with inventory and status data, there must be a so-called identifier in the entries of these lists in order to link the related elements, eg.:

    for sid, data in tablespaces:
        inv_node.append({"sid": sid, "INV-KEY": data[INV-DATA]})
        status_node.append({"sid": sid, "STATUS-KEY": data[STATUS-DATA]})

Both nodes, 'inv_node' and 'status_node', use 'sid' as an identifier. Thereby Check_MK knows which entries of both lists belong together. If one identifier is not sufficient you can add more, eg.:

    for sid, data in tablespaces:
        inv_node.append({"sid": sid, "name" : name, "INV-KEY": data[INV-DATA]})
        status_node.append({"sid": sid, "name": name, "STATUS-KEY": data[STATUS-DATA]})

Now Check_MK uses 'sid' and 'name' as identifier. These identifier have to exist in all entries of both lists. Otherwise there's no possibility to link these entries.

To make Check_MK know which inventory plugins calculate status data an additional attribute 'has_status_data' must be added which has the value 'True'.

inv_info['oracle_tablespaces'] = {
    "inv_function": inv_oracle_tablespaces,
    "has_status_data": True,
}

To the list of all Werks