Zabbix average calculation

[JeroenE]

Hello Zabbix Community!

I'm hoping to get some insights into an issue I've encountered with uptime calculation in Zabbix.

Here's the situation: we have a Zabbix server monitoring a target server with an ICMP ping check running every minute (icmpping[,5], Interval 1m, History 90d, Trends 365d).
Over a 30-day period (43,200 minutes), Zabbix collected 43,196 values (March 1 00:00:00 to March 31 00:00:00)

• 43,194 values indicated the server was up
• 2 values indicated the server was down

When I manually calculate the uptime percentage using the formula: Uptime % = (Up values / Total values) * 100 = (43,194 / 43,196) * 100 = 99.995%
However, Zabbix is reporting an average uptime of 99.86% for the same period. I suspect the discrepancy might be related to Zabbix's averaging methodology over long time periods.

I've reviewed the documentation and forum posts, but haven't found a clear explanation for this specific scenario. My questions are:

1. What could be causing the difference between the manual calculation (99.995%) and Zabbix's reported uptime (99.86%)?
2. How does Zabbix handle missing data points in its uptime calculation?
3. Does Zabbix use any approximations or optimizations when calculating average uptime over long periods (e.g. 30 days)?
4. Are there any server settings or configuration options that could influence the calculation?

Any insights or guidance on resolving this discrepancy would be greatly appreciated. If I can understand and explain to my peers how Zabbix comes to the average value, i would much appreciate it!

Thank you in advance for your help!

[JeroenE]

Does anyone have any idea how averages are calculated? Or can someone steer me in the right direction?

[cyber]

Does Zabbix use any approximations or optimizations when calculating average uptime over long periods (e.g. 30 days)?

I think in some cases yes... I cannot remember the details, but I think I have stumbled on that in the past...

[JeroenE]

It appears to be because a graph is created with trends, not history, when the graph is showing more than 24h periods. The avg is calculated on the trend, not all values, this is why its lower.

Unfortunately i cannot confirm this by the documentation or other, it's just what i am observing. If anyone can help me verify my claim i would love the input!