archlinux, postgresql 13.3, timescaledb 2.3.1

the problem is reproduced better, when using 15-20 hosts for test

Hello!
This problem was fixed in 5.4.3.
Upgrade, try again and report your findings.

Good luck.
Andrey

same problem in 5.4.3

Hello again!
Still can not reproduce.
Added only snmp interface to host. Added snmp and icmp checks. Bringed down host for 10 minutes. After host up all metrics works again.
What am i missing?

Problem reproduced, when host going down (not up!) I show you 2 screenshots, pay attention for item timestamp.

No problem (version 5.2.6): "0" value every 1 min.

Problem (v.5.4.3): First "0" value received after 5min! (but it does 1min).  Thus, the "host unavailable by icmp ping" message arrives much later than it should. 

Here is my tests result. Table header contains keys, as items can differ from template to template. Module templates mentioned before not available in 5.4 (only if tagged along after upgrades from previous version)
Changed interval to scheduled (once per 5 seconds) to get clear picture.
There are icmpping item, then internal item to check interface availability(internal check) and 2 snmp item with some OIDs

Immediately after router was shutdown icmpping returns 0 and no snmp data is available. It does not matter how much more items added to that host or how much interfaces is there. Always result is the same.

Same with multiple hosts:

Do I understand correctly that in your screenshot, zabbix[host,snmp,available] data item continues to return "1" after the router has been turned off?

This means, that the host interface will not go to Unreacheble state, and icmpping check will work correctly.

In my case snmp available check changing value to 0, and host interface went to unreachable state.

In my case SNMP interface goes to unreachable state in a minute, but this does not influence icmp checks.

I confirm this on my test installation with current GIT rev 5.4.4rc1
Also confirmed from 2 production installations on 5.4.3

It's very simple to reproduce: have host with agent interface, have 2 items: agent and simple one, 1 minute update interval.
Have both items collecting values properly.
Stop agent (it get unavailable) and observe how it stops the simple check collection.

A few examples:

Here? after long time waiting, the simple check starts randomly? to pool again, but later it may have gaps again.
Host was turned off:

1740899:20210825:225840.472 Starting Zabbix Server. Zabbix 5.4.4rc1 (revision {ZABBIX_REVISION}).
...
1740924:20210825:230348.669 Zabbix agent item "agent.version" on host "W7" failed: first network error, wait for 15 seconds
1740929:20210825:230406.690 Zabbix agent item "agent.version" on host "W7" failed: another network error, wait for 15 seconds
1740929:20210825:230424.692 Zabbix agent item "agent.version" on host "W7" failed: another network error, wait for 15 seconds
1740929:20210825:230442.695 temporarily disabling Zabbix agent checks on host "W7": interface unavailable

Timestamp	Value
2021-08-25 23:19:30	0
2021-08-25 23:18:21	0
2021-08-25 23:17:15	0
2021-08-25 23:03:29	0
2021-08-25 23:02:29	1
2021-08-25 23:01:29	1

Here host was up, but agent was stopped. I was waiting for 15 minutes but simple check did not collect values this time. Not sure why:

1740926:20210825:232545.850 Zabbix agent item "agent.version" on host "W7" failed: first network error, wait for 15 seconds
1740929:20210825:232600.221 Zabbix agent item "agent.version" on host "W7" failed: another network error, wait for 15 seconds
1740929:20210825:232615.223 Zabbix agent item "agent.version" on host "W7" failed: another network error, wait for 15 seconds
1740929:20210825:232630.226 temporarily disabling Zabbix agent checks on host "W7": interface unavailable

2021-08-25 23:25:29	1
2021-08-25 23:24:29	1
2021-08-25 23:23:29	1

The same, with 127.0.0.1 IP:

1740926:20210825:234028.962 Zabbix agent item "agent.version" on host "it0" failed: first network error, wait for 15 seconds
1740929:20210825:234043.348 Zabbix agent item "agent.version" on host "it0" failed: another network error, wait for 15 seconds
1740929:20210825:234058.350 Zabbix agent item "agent.version" on host "it0" failed: another network error, wait for 15 seconds
1740929:20210825:234113.352 temporarily disabling Zabbix agent checks on host "it0": interface unavailable

2021-08-25 23:40:06	1
2021-08-25 23:39:06	1
2021-08-25 23:38:06	1
2021-08-25 23:37:06	1

Maybe it would be better to simply remove the binding to the interface at a simple check item? There aren't such binding at database monitor or script item types.

Fixed in pull request feature/ZBX-19712-5.4

Fixed in

• pre-5.4.5rc1 5f81ea74be9
• pre-6.0.0alpha3 (master) dcfdbdc63fc

Updated documentation:

• Added to Known issues.