We are reporting an issue with Zabbix 6.0.15.

We use RHEL-9 and the official Zabbix packages. After updating from 6.0.14 to 6.0.15 all items, collected over SNMP that are either Counter32 or Counter64 stopped working. This is not vendor-dependent, e.g. for us it affects equally Cisco and HP devices. Remaining SNMP items (mostly gauges) still work properly. Example: on the same interface ifInOctets no longer works, but ifAdminStatus does work. Other tools like snmpget fetch the values of these problematic items properly (both for Counter32 via SNMP v1 and for Counter64 via SNMP v2c); just in case I inspected a traffic dump when snmpget runs and the reponse does include proper tag for the value type, e.g. 0x46 for Counter64. The problematic items are still discovered properly (no message saying they are not discovered any more). Zabbix does not report these items as unsupported or any other failure in their data processing. There are no related messages in the Zabbix server log. We checked the database entries for one such item and found that the last row for this item was written to the history_uint table precisely at the time of the Zabbix version update; before it there was an entry every 5 minutes (at which interval we collect this data) and after this time there were no records for this item in the database.

Reverting to 6.0.14 restores the proper collection of the said items. 

I checked the Release Notes of 6.0.15 for anything suspicious, but the only mention of SNMP was in ZBX-22026; as much as I'd like to blame it, the three-line patch seems fairly innocent. I could not find anything similar reported. 

I just spotted 6.0.16 was out, but we don't have it yet (our lifecycle manage delays updates to production by up to two weeks) - however, its Release Notes do not mention anything similar been fixed.

We'd appreciate if anybody with more knowledge of the changes in this release could consider the report.