Maximum size of text values (with proxy !)

[Albéric de Pertat]

Hi everyone

I am currently testing the 4.0.0alpha9 of Zabbix, which is great so far. However I stumbled upon the following problem : I am trying to use the dependent item feature to query a big (200kb) chunk of XML served over HTTP (FYI it's the Bind 9 statistics server). For this I am using the web.page.get item through the Zabbix agent. The documentation states that I should be able to process 512kb but only 64kb will be kept in the history table. This should allow me to preprocess my XML with some XPath requests and create my dependent items. However Zabbix truncates the HTTP request before the first preprocessing step (removing the HTTP headers with a regex).

I am getting the web.page.get value through a Zabbix proxy because my name server is in a DMZ. As proxies do not run the preprocessing workers and only servers do, am I right in assuming that my item output will be stored in the history_text table of the proxy, which will truncate it ?

I have already considered a number of workarounds but it seems to me that this behaviour is unexpected and as far as I can tell, undocumented yet.

Thanks

[dimir]

Unfortunately Zabbix agent cuts any received value to 64 Kb:

$ curl -s http://127.0.0.1/tmp/bigtext.txt | wc -c
308543

$ bin/zabbix_get -s 127.0.0.1 -p 11339 -k vfs.file.size[/var/www/html/tmp/bigtext.txt]
308543

$ bin/zabbix_get -s 127.0.0.1 -p 11339 -k vfs.file.contents[/var/www/html/tmp/bigtext.txt]
ZBX_NOTSUPPORTED: [COLOR=#FF0000]File is too large for this check[/COLOR].

$ bin/zabbix_get -s 127.0.0.1 -p 11339 -k web.page.get[127.0.0.1,/tmp/bigtext.txt] | wc -c
[COLOR=#FF0000]65536[/COLOR]