Data Explorer vs. Host Overview – Difference in Host Availability % for AIX Servers

[Sohel_Rashid]

Hi Dynatrace Community,

We are observing a significant difference in the Host Availability % between Data Explorer and the Host Overview page for some AIX servers.

Environment

• OS: AIX
• Hosts: UPI DB1 / UPI DB2
• Time period: September
• Dynatrace: Managed environment

Issue

When we check the availability for the same hosts and the same time period:

• Data Explorer: Host Availability is showing approximately 35.3%
• Host Overview: Host Availability is showing approximately 99%

This is a significant difference, even though we are checking the same hosts and September time period.

What we would like to understand

Could someone please clarify:

1. What is the exact metric/key and calculation used by the Host Overview page to calculate Host Availability %?
2. Is the availability shown in Data Explorer calculated differently from the Host Overview page?
3. Could this difference be related to:
   • Metric aggregation
   • Availability event calculation
   • Timeframe/timezone
   • Host monitoring state
   • Missing telemetry/data points
   • AIX-specific monitoring behavior
4. Which value should be considered the correct Host Availability % for monthly reporting/SLA purposes?

We would also appreciate it if someone could provide the recommended metric selector/DQL query to reproduce the same availability percentage shown on the Host Overview page.

I can share screenshots, the Data Explorer configuration/metric selector, host details, and the exact timeframe if required.

Thanks in advance for your help.

[sujit_k_singh]

Hi @Sohel_Rashid 

This is a well-known discrepancy in Dynatrace. Let me answer your question one by one

1. 1. What is the exact metric/key and calculation used by the Host Overview page to calculate Host Availability %? 
2. The Host Overview page does not use a raw timeseries metric for its availability display. it calculates availability based on problem/availability events recorded against the host entity. This tends to produce high availability numbers (~99%) because it only counts explicit DOWN events.
3. 2. Is the availability shown in Data Explorer calculated differently from the Host Overview page?
4. Yes, It uses the dt.host.availability timeseries metric, which is reported every minute based on OneAgent heartbeat state. If OneAgent is not actively sending data (any non-UP state), those minutes count as unavailable.
5.  
6. 3. Could this difference be related to:
   • Metric aggregation - Yes, partially. If you use avg instead of max(rollup: sum) in Data Explorer, results will be skewed. The correct aggregation for availability % is max with rollup: sum, then normalize against interval length.
   • Availability event calculation - Yes, this is the primary cause.
   • Timeframe/timezone - Possible but I have to check this point separately 🙂
   • Host monitoring state - Yes, significant contributor
     
   • Missing telemetry/data points - Yes,  directly causes the gap. Every minute where no heartbeat is received registers as a non-UP data point. A 35% availability figure means maximum number of minutes were non-UP — this points to a persistent agent issue, not just occasional gaps.
   • AIX-specific monitoring behavior - Yes, AIX is a known factor. AIX OneAgent has historically had more telemetry interruptions compared to Linux/Windows
7. 4. Which value should be considered the correct Host Availability % for monthly reporting/SLA purposes? It depends on how your SLA is defined but the metric-based (Data Explorer) value is more rigorous.
   Example:
   If SLA like "Host was not in a detected DOWN/problem state" then use Host Overview (~99%)
   If SLA like "OneAgent was actively monitoring and reporting every minute" then use Data Explorer (~35%)

Thanks,

Sujit