Table of Contents
11 Unreachable/unavailable host interface settings
Overview
Several configuration parameters define how Zabbix server should behave when an agent check (Zabbix, SNMP, IPMI, JMX) fails and a host interface becomes unreachable.
Unreachable interface
A host interface is treated as unreachable after a failed check (network error, timeout) by Zabbix, SNMP, IPMI or JMX agents. Since Zabbix 6.2.0, active Zabbix agent checks also affect interface availability. When active checks become unavailable, they contribute to the overall availability status of the agent interface.
From the moment an interface becomes unreachable, UnreachableDelay defines how often it is rechecked using one of the items (including LLD rules). These rechecks are performed by unreachable pollers (or IPMI pollers for IPMI checks). By default, the next check is in 15 seconds.
Checks performed by asynchronous pollers are not moved to unreachable pollers.
In the Zabbix server log unreachability is indicated by messages like:
Zabbix agent item "system.cpu.load[percpu,avg1]" on host "New host" failed: first network error, wait for 15 seconds
Zabbix agent item "system.cpu.load[percpu,avg15]" on host "New host" failed: another network error, wait for 15 secondsThe log messages specify the exact item that failed and its type (Zabbix agent).
The Timeout parameter will also affect how early an interface is rechecked during unreachability. If Timeout is set to 20 seconds and UnreachableDelay to 30 seconds, the next check will be in 50 seconds after the first attempt.
The UnreachablePeriod parameter defines the total duration of the unreachability period. By default UnreachablePeriod is 45 seconds. This value should be several times larger than UnreachableDelay to ensure that an interface is rechecked multiple times before being marked as unavailable.
An internal item, zabbix[host,active_agent,available], allows monitoring the availability of active checks in unreachable scenarios.
Switching interface back to available
When the unreachability period is over, the interface is polled again, decreasing priority for item that turned the interface into unreachable state. If the unreachable interface reappears, the monitoring returns to normal automatically:
Once interface becomes available, the host does not poll all its items immediately for two reasons:
- It might overload the host.
- The interface restore time is not always matching planned item polling schedule time.
So, after the interface becomes available, items are not polled immediately, but they are getting rescheduled to their next polling round.
Unavailable interface
After the UnreachablePeriod ends and the interface has not reappeared, the interface is treated as unavailable.
In the server log it is indicated by messages like these:
and in the frontend the host availability icon goes from green/gray to yellow/red (the unavailable interface details can be seen in the hint box that is displayed when a mouse is positioned on the host availability icon):
The UnavailableDelay parameter defines how often an interface is checked during interface unavailability.
By default it is 60 seconds (so in this case "temporarily disabling", from the log message above, will mean disabling checks for one minute).
When the connection to the interface is restored, the monitoring returns to normal automatically, too: