Queue is filling up SNMP and IPMI Items - since upgrade from 3.2 to 3.4

[rglade]

we have the problem that the queue for SNMP and IPMI entries keeps filling up. Interestingly, this only happens with certain items with certain items. Mostly it's items with a longer interval such as 6 hours or one day.
The problem problem occurs so massively only since the conversion from 3.2 to 3.4.
Other items will be processed cleanly in time! After a restart of Zabbix, it may also be that the items are retrieved for a while (quite possibly for two days) - but at some point this process seems to get unsteady. Originally, I suspected an overload of the Zabbix server and outsourced the requests to a proxy. Apparently this has make the problem not better. Even the adjustment of the configuration values (on proxy and server) has not make anything better:

 ConfigFrequency=3600 StartPollers=60 StartIPMIPollers=10 StartPollersUnreachable=200 StartTrappers=50 StartPingers=30 StartHTTPPollers=5 StartVMwareCollectors=5 VMwareFrequency=120 VMwarePerfFrequency=120 VMwareCacheSize=16M CacheSize=128M HistoryCacheSize=128M HistoryIndexCacheSize=32M ExternalScripts=/usr/lib/zabbix/externalscripts FpingLocation=/usr/bin/fping StartPingers=25 Timeout=30 TrapperTimeout=120

In fact, I've found that the affected devices will not answer for a short time. This is also logged. But then, when answers are sent again, the queue does not seem to be processed again. As already mentioned, the problem with SNMP and IPMI trembles. I hope that somebody had a similar problem and maybe found a solution. Maybe there is also the possibility to force the processing of the queue? VG

[LenR]

What version of 3.4? What is your NVPS load?