Which SNMP Walk preprocessing is faster, Regex or JSONPath

[markfree]

I have some devices that discover many items each.
I'm refactoring their template so that the devices can benefit from the SNMP Walk feature.

The parent "Walk" item finds many SNMP OIDs and its dependent items use preprocessing to identify the item values.

As I see it, Regex should work fine for each dependent item.
However, if I map the SNMP walk parent item to JSON, I could benefit from the JSONPath preprocessing.

So my question is, which preprocessing do you guys think would be best:

• Main item with no preprocessing and simple regex for each dependent item?
• Or, main item with "SNMP walk to JSON" preprocessing, and JSONPath preprocessing for each dependent item?

PS: I'm thinking of thousands of items.

[cyber]

Regex will go over whole "file" line by line (until match). Jsonpath walks the path... Thus I would say jsonpath is faster...

[markfree]

I guess so...
Is there a way to measure each one?

I'll try to debug these preprocessing steps.

[cyber]

I have no idea, how to benchmark it...

[aigars.kadikis]

Starting with 7.0.17, 7.4.1 there is a runtime command to ask about which itemid bursts the preprocessing:
zabbix_proxy -R diaginfo=preprocessing

https://support.zabbix.com/browse/ZBXNEXT-10119

I think to make benchmarking who will win JSONPath VS regEx, we can create a master item "HTTP agent", "Zabbix trapper", "Script" (and hardcode input) and 2 dependent items.
Set the 3 items to be only items on a single proxy.

Here is minimal (active) proxy config (can run in parallel of existing proxy)

echo "
DBName=/tmp/min_zabbix_proxy.sqlite3
Hostname=minimal
LogType=console
PidFile=/tmp/minimal_zabbix_proxy.pid
Server=aigars.zabbix.cloud
StartAgentPollers=0
StartBrowserPollers=0
StartDBSyncers=1
StartDiscoverers=0
StartHTTPAgentPollers=0
StartHTTPPollers=0
StartODBCPollers=0
StartPingers=0
StartPollers=1
StartPollersUnreachable=1
StartPreprocessors=1
StartSNMPPollers=0
StartTrappers=0
" | sudo tee /root/minimal_zabbix_proxy.conf

to run in foreground:

zabbix_proxy -c /root/minimal_zabbix_proxy.conf -f

In another terminal see stats of itemIDs

zabbix_proxy -c /root/minimal_zabbix_proxy.conf -R diaginfo=preprocessing

I tried to benchmark myself with this Proof of Concept host (not template):

{
    "zabbix_export": {
        "version": "7.0",
        "host_groups": [
            {
                "uuid": "e34c64a05273412ba868d8b04b812089",
                "name": "tempo"
            }
        ],
        "hosts": [
            {
                "host": "regEx VS JSONPath",
                "name": "regEx VS JSONPath",
                "monitored_by": "PROXY",
                "proxy": {
                    "name": "minimal"
                },
                "groups": [
                    {
                        "name": "tempo"
                    }
                ],
                "items": [
                    {
                        "name": "input",
                        "type": "SCRIPT",
                        "key": "input",
                        "delay": "1s",
                        "value_type": "LOG",
                        "trends": "0",
                        "params": "return '[{\"name\":\"hello\",\"surname\":\"world\"}]';",
                        "timeout": "1s"
                    },
                    {
                        "name": "jsonpath",
                        "type": "DEPENDENT",
                        "key": "jsonpath",
                        "delay": "0",
                        "value_type": "LOG",
                        "trends": "0",
                        "preprocessing": [
                            {
                                "type": "JSONPATH",
                                "parameters": [
                                    "$[0].name"
                                ]
                            }
                        ],
                        "master_item": {
                            "key": "input"
                        }
                    },
                    {
                        "name": "regex",
                        "type": "DEPENDENT",
                        "key": "regex",
                        "delay": "0",
                        "value_type": "LOG",
                        "trends": "0",
                        "preprocessing": [
                            {
                                "type": "REGEX",
                                "parameters": [
                                    "name...([^\"]+)",
                                    "\\1"
                                ]
                            }
                        ],
                        "master_item": {
                            "key": "input"
                        }
                    }
                ],
                "inventory_mode": "AUTOMATIC"
            }
        ]
    }
}

After waiting 3m and executing master item every 1s, the RegEx itemid appiered in list which makes JSONPath the winner.