NodeID integer only?

[martin.marcher]

Hi,

are the node IDs in distrubuted setups integer only or can we just use a UUID that we generate?

[Alexei]

NodeIDs are integers, no UUID can be used.

[martin.marcher]

NodeIDs are integers, no UUID can be used.

Is there a reason for that?

I've been thinking about it and don't quite understand this. We need:

• unique hostname for active checks
• unique nodenames (on the server)

How come?

With a UUID you could always have it generated anyway and kill a source of error in the configuration.

To me that sounds like a good idea to use UUIDs instead of numeric NodeIds (and hostnames in the agent config)

[Alexei]

Is there a reason for that?

Sure there is! It is because we have a limit of 1K of nodes maximum.

[martin.marcher]

Sure there is! It is because we have a limit of 1K of nodes maximum.

Are you saying that I can't have NodeIds starting at 1000?

Because I can't see how with an SQL backend a numeric NodeID correlates to the number of tuples in a table.

Does that mean I can't have:

• NodeId 1
• NodeId 950
• NodeId 1050

Seems strange to me.

[Alexei]

Are you saying that I can't have NodeIds starting at 1000?

Of course you cannot. Check ZABBIX Documentation. The ID is in range of 0-999 only.

Hmm, a correlation between NodeID and the number of tuples?!

[martin.marcher]

I'm thinking of something that pushs/pulls the config from the master (like with proxies if I understand "16.2.Proxy v.s. Node" correctly) - of course those would need a unification before that.

zabbix=# \d hosts
                                     Table "public.hosts"
       Column       |          Type          |                    Modifiers
--------------------+------------------------+-------------------------------------------------
 hostid             | bigint                 | not null default 0::bigint
 proxy_hostid       | bigint                 | not null default 0::bigint
 host               | character varying(64)  | not null default ''::character varying
 dns                | character varying(64)  | not null default ''::character varying
 useip              | integer                | not null default 1
 ip                 | character varying(39)  | not null default '127.0.0.1'::character varying
 port               | integer                | not null default 10050
 status             | integer                | not null default 0
 disable_until      | integer                | not null default 0
 error              | character varying(128) | not null default ''::character varying
 available          | integer                | not null default 0
 errors_from        | integer                | not null default 0
 lastaccess         | integer                | not null default 0
 inbytes            | bigint                 | not null default 0::bigint
 outbytes           | bigint                 | not null default 0::bigint
 useipmi            | integer                | not null default 0
 ipmi_port          | integer                | not null default 623
 ipmi_authtype      | integer                | not null default 0
 ipmi_privilege     | integer                | not null default 2
 ipmi_username      | character varying(16)  | not null default ''::character varying
 ipmi_password      | character varying(20)  | not null default ''::character varying
 ipmi_disable_until | integer                | not null default 0
 ipmi_available     | integer                | not null default 0
 snmp_disable_until | integer                | not null default 0
 snmp_available     | integer                | not null default 0
Indexes:
    "hosts_pkey" PRIMARY KEY, btree (hostid)
    "hosts_1" btree (host)
    "hosts_2" btree (status)
    "hosts_3" btree (proxy_hostid)

Insert a node_hostid and you are done with:

select hostid, proxy_hostid, node_hostid, host from hosts;
select count * from from hosts where node_hostid <> '';

one could than use arbitrary proxy/node IDs in the config.

[Alexei]

Please do more code research (hint, structure of any ID in the database) or visit "ZABBIX for Large Environments" training course. The limit of 1K of nodes is much more fundamental, it has nothing to do with a number of records, type or length of NodeID, whatever. Structure of the IDs (like itemid, hostid, etc) is what really matters.

[martin.marcher]

...or visit "ZABBIX for Large Environments" training course....

I've been thinking about that actually.

Right now I'm fighting with OpenVZ and the ZBX_TCP_READ problem some people are having.

Well I'll see I've been using zabbix in a small env with about 30 hosts in my last company and now I'm trying to get my colleagues into zabbix where we currently have ~200 instances in AT, NL and US.

I don' remember that problem from 1.4 but I guess it's more a problem between keyboard and chair