Production

Is it possible to determine which exact process is leaking please and if disabling some specific item types help with the issue ?
ps -aux should help

there are no defunct'd processes, the results from ps -aux look fine.

Can you please share results of ps -aux | grep zabbix

here are the results. of ps -aux|grep zabbix

results

Also please provide the following information:
SQL query from Zabbix proxy:

select count(*), type from items group by type;

Package dependencies. For example RHEL like systems command:

yum deplist "zabbix-proxy*"

mysql> select count, type from items group by type;
--------------+

count

type

--------------+

--------------+
6 rows in set (0.02 sec)

dep_results

I guess the problem is with a poller

zabbix    2664  0.0  0.1 2850604 16788 ?       S    08:43   0:02 /usr/sbin/zabbix_proxy: poller #2 [got 0 values in 0.000015 sec, idle 1 sec]

Does disabling simple checks help with the issue ? Are you using any loadable modules ?

Way too many processes started on the proxy to begin with. You have 350 pollers for 287 agents. Same goes for caches and DB syncers. Lower the settings to reasonable ones and see how it does then. 

Related "memory leak" reported in ZBX-15996

Closing a a duplicate of ZBX-15996, please continue discussion there

The ticket you stated that is a dup to this, is on a different version.  also, I have the same conf setting on my other proxies (22 of them) and do not see the memory leak on any of them.. what is the guide line to setting the conf values, example x number of items needs x number of pollers, is there a doc with that info.

Can I get an update on this issue?  I am starting to see memory lose in a few of my other proxies now, just not as fast as the proxy that I open the ticket on..  you can see the slow decline of memory and then the reboot.. 

Does disabling simple checks help with the issue ? Are you using any loadable modules ? Could not reproduce issue locally.

The proxy that is losing memory faster than the others, most of the checks preformed are windows eventlog and discovery that collects 9 item values.  the event log checks are active agent and everything else is agent.

287 agents reporting to the proxy, 19780 items.

there is only 1 simple check, ping

Is it possible to move all other checks except eventlog to other Zabbix proxy and seeng if leaving only eventlog causes memory leak ?

I can not easily move the host around, the proxies are there with specific firewall rules to specific hosts.  

Can you please do

select count(*), type from items group by type;

On Zabbix proxy that leak, judging by report it's either simple check or snmpv2

mysql> select count, type from items group by type;
--------------+

count

type

--------------+

--------------+

Thanks ! In the other ticket it mentions

count | type 
-------+------
22229 |    4   snmpv2
  5150 |    6   snmpv3
  3706 |    3   simple check
  3558 |   10  external check
  3560 |    5   zabbix internal

So if true then it's either 4 or 3 that is Simple check or SNMPv2 agent.
It says that you have 580 simple checks and 462 SNMPv2 agent items, you said that you have only one simple check and it is ping, could you please double check ?