When new VMware sevice is configured the collector tries to retrieve all events. This is done with batches by 10 events and can take long time. Debug log contains a lot of

 14153:20181009:195424.262 In vmware_service_read_previous_events()
 14153:20181009:195424.278 End of vmware_service_read_previous_events():SUCCEED
 14153:20181009:195424.279 In vmware_service_parse_event_data() last_key:0
 14153:20181009:195424.288 End of vmware_service_parse_event_data() parsed:10
 14153:20181009:195424.288 In vmware_service_read_previous_events()
 14153:20181009:195424.303 End of vmware_service_read_previous_events():SUCCEED
 14153:20181009:195424.303 In vmware_service_parse_event_data() last_key:0
 14153:20181009:195424.313 End of vmware_service_parse_event_data() parsed:10
 14153:20181009:195424.313 In vmware_service_read_previous_events()
 14153:20181009:195424.330 End of vmware_service_read_previous_events():SUCCEED
 14153:20181009:195424.330 In vmware_service_parse_event_data() last_key:0
 14153:20181009:195424.340 End of vmware_service_parse_event_data() parsed:10

which can be misleading - letting people think that collector entered in endless loop.

What could be done:

• improve debug logging to show the progress of event retrieval
• investigate if the batch limit of 10 events could be safely increased. Actually it makes sense to use smaller batches to update events (for existing hosts) and larger batches to download all events (for new hosts).
• either add 'skip' option or make new VMware monitoring hosts skip the old events by default
• add internal timeout so event downloading does not take too long. This way it would take serveral updates to catch up with events, but it would not block vmware collector. Cannot do this - while collector will download all events Zabbix would use only the first downloaded batch.