The graph shows a short spike for a garbage collection issue

Problem:

We identified a garbage collection (GC) issue, and described it lasting for 3 minutes. However, the graph just shows a short spike. Why is that ?

Solution:

The GC too high issue is created when GC goes over 10% over a 1 minute tumbling window. After the 1 min, the issue expires in 2 minutes. That means 1 minute after we detect it we trigger the issue, starting it when we first detected it and then in 2 minutes expire the issue, thus the 3 minutes. The graph won't show GC too high for three solid minutes. If it did that, the issue would be open for longer and would likely be at least 5 minutes long.

 

- average garbage collection time of ~3s in a the 5s rollup time window. So this means ~15s garbage collection time at this spike, i.e. ~15s spent garbage collecting within 5s.
 - The rule triggers when more than 6s are spent garbage collecting within a minute, i.e. 10% GC time.
 - The actual GC time the customer had within that minute was therefore ~25%.
 - We generate a warning when GC time exceeds 10%. 
 - We generate a critical when GC time exceeds 50%.

Have more questions? Submit a request

Comments