With Agent interface all work fine. If only SNMP interface is specified in a host the macro expands according documentation.
http://www.zabbix.com/documentation/2.0/manual/appendix/macros/supported_by_location
"In the context of remote command execution, GUI scripts, item key's parameters, and interface IP/DNS fields, only the main Agent interface will be considered as the source of information."
I know this problem.
I do not like current behavior (as described in documentation) because to use GUI scripts for a SNMP-based device I have also add Agent interface to already existing and used (only) SNMP interface. A real case described in the ZBXNEXT-1089
Maybe it could be reviewed ?
Rich, do not forget about configuration cache update interval.
after a bunch of testing, yes, configuration caching confused me 
unless i'm forgetting something, current behaviour doesn't seem to make much sense - this makes frontend scripts unusable for hosts with snmp interface only, unless "dummy", unused agent interface is added as well.
i'd probably go for agent-snmp-jmx-ipmi, to be the same ordering as in the frontend
<zalex> I recall I was also mislead because of using server cache (for interfaces) for macros expending when I created the ZBXNEXT-1089
So, this is an unclear problem already for two people and it's serious indicator
Maybe server cache should not be used here (for GUI scripts) ?
For example it's already done for disabled host (ZBX-4889)
Users can change interfaces setting (add/delete, change IP) and very fast try to use GUI scripts (with macros) and it's absolutely not clear why they do not work or work as not expected.
From forum I recall that many users (even experienced) did not know that zabbix scripts actually executed by server daemon, but not by the fronted itself.
So, I support Rich's point of view (agent-snmp-jmx-ipmi) and I suggest to not use server cache to expanding macros in GUI scripts.
<richlv> not sure about skipping the cache, performance is the king
google translate
---------------------------------------------
The device was found on snmp (autodiscovery), respectively, the agent interface he does not - but there is an element of data a_est_li_eto_chudo_na_karte ["
{HOST.CONN}"]respectively, the work was broken. manually add the interface agent incorrectly - what shall we do with it? Logical external audits performed on any interface ip if it is.
original
---------------------------------------------
Устройство нашлось по snmp (автообнаружение) соответственно интерфейса агента у него нет - но есть элемент данных а_есть_ли_это_чудо_на_карте["{HOST.CONN}
"]
соответственно работа сломалась. вручную добавлять интерфейс агента неправильно - что будем делать с этим? Логичней внешние проверки выполнять на ип любого интерфейса если он есть.
ZBXNEXT-1293 is related
Additional forum thread (in Russian) http://www.zabbix.com/forum/showthread.php?t=27075
ah, no need to file a new ticket...
We used to do this via a template on a couple of hundred devices, simple external check with a custom script in 1.8.12:
Template_X_Poiprouter: PoipLocalIkePeer PoipLocalIkePeer.sh["
{HOST.CONN}",] 86400 14 External check
Now we migrated to 2.0.1, all the items are stuck in the queue - big issue.
Thanks!
Fixed in the development branch svn://svn.zabbix.com/branches/dev/ZBX-4976
(1) it seems we need to see updated documentation do know how it should work now
<zalex> ohh, SVN log message in the dev branch is very useful, I'm copying it here:
........S. ZBX-4976 improved
, {HOST.IP) and
{HOST.DNS} macros substitution for global scripts
Before this change:
- macros are replaced with values from the host's main "Zabbix agent" interface
- macros are replaced with UNKNOWN if such interface isn't present
After this change: - macros are replaced with values from the host's main interfaces in the following order:
- "Zabbix agent" interface
- SNMP interface
- JMX interface
- IPMI interface
- macros are replaced with UNKNOWN if such interfaces aren't present
<richlv> should be documented in http://www.zabbix.com/documentation/2.0/manual/appendix/macros/supported_by_location and the note below the table updated (mentioning which version changes the behaviour, probably 2.0.3)
<Sasha> updated documentation. Please review.
<richlv> docs seem to be good, CLOSED
the dev branch successfully tested. works as described above.
Successfully tested!
ZBX-3619 talks about more global lack of consistency in this area
Fixed in pre-2.0.2 r29111 and pre-2.1.0 (beta) r29112.