Documentation https://www.zabbix.com/documentation/2.4/manual/discovery/network_discovery states:

Note that one discovery rule will always be processed by a single discoverer process. The IP range will not be split between multiple discoverer processes.

I don't think anything can be changed in this regard.

Especially taken into account that recently we extended syntax of supported IP ranges - now it may be too complex to be split on parts:
https://www.zabbix.com/documentation/2.4/manual/introduction/whatsnew244#extended_ip_range_option_in_network_discovery

I believe that ZBXNEXT-2721 should resolve current troubles.

how it work for multiple discovery rules ? would all other rules also processed by same single discoverer?

rubal033, while the documentation does not explicitly say that, if there are multiple discoverer processes, they will split the rules between themselves (but a single rule will never be split).

this feature request asks to split rules among multiple discoverer processes.
potentially, discoverers could also work on multiple rules in parallel, but that will depend on the implementation detail.

I thought about how to divide the range with Zabbix API and register it.

python splite_drule.py <url> <user> <password> <iprange> <split> <json file>

<url>       : http://192.xxx.xx.xxx/zabbix/api_jsonrpc.php
<user>      : admin
<password>  : zabbix
<iprange>   : 192.168.0.0/16
<split>     : 19
<json file> : /tmp/drule.json

drule.json

{
    "jsonrpc": "2.0",
    "method": "drule.create",
    "params": {
        "name": "test",
        "aiprange": "iprange",
        "delay": "86400",
        "dchecks": [
            {
                "type": "12",
                "uniq": "0"
            },
            {
                "type": "11",
                "key_": "1.3.6.1.2.1.1.1",
                "ports": "161",
                "snmp_community": "public",
                "uniq": "0"
            }
        ],
        "status" :"1"
    },
    "auth": "auth",
    "id": 1
}

If Zabbix have ability add address list in one rule, then Zabbix must handle this list as fast as usual "one subnet" rule.  Otherwise the ability to create lists is simply cosmetic, but not usable in production.

At least the list should be splitted by regexp by comma/space and distributed to all discoveres. Best solution: calculate list address space size and split subnets to all discoverers. Medium: split big subnets to /24 and distribute

Available in versions:

• 7.0.0alpha2 (master) 14d9969905b, cb3805d3fdc, a753a939531

Updated documentation:

• What's new in 7.0
• Upgrade notes
• Network discovery
• Server processes (new processes)
• Proxy processes (new processes)
• Internal items (new discovery queue item)
• Internal items (process item updated)
• StartDiscoverers server parameter (updated)
• StartDiscoverers proxy parameter (updated)
• API Discovery rule object (new property)
• API drule.create (updated)
• API drule.get (updated)
• API Appendix 2. Changes from 6.4 to 7.0