NameNode Heap Usage (weekly) alert - I get these periodically
Critical is set to 60%, warning at 50% minimum heap set to 1000M. The variance for this alert is 504MB which is 66% of the 7560MB average (4536MB is the limit).
Is this a concern? If so, what should I be looking at?
There are two variance of "NameNode Heap Usage" alert (Daily & Weekely), which is a service-level alert and it is triggered if the NameNode heap usage deviation has grown beyond the specified threshold within a given period.
it just shows the deviation of heap usages if it gooes beyond the mentioned throshold. That helps us in planning in advance if we need to increase the NameNode heap (if we have set the threshold properly).
The default growth rate in heap for CRITICAL is set to 50% and WARNING is set to 20%. This varies based on the usage an type of your cluster and HDFS so this value users can adjust based on their requierment and based on the nature and heavy use.
So in your case if you think that based on the usage this much growth is not expected OR in future if you think that your NameNode will require more Heap then please plan to increase the Heap of the NameNode (because as much as the data will grow in HDFS, it's metadata size will also grow on NameNode heap)