How reliable are indexer's statistics?

Couchbase Server
1

On CB 6.6.2 EE’s web console, I’m giving up on getting a solid picture of possibily running out resources from figures in

  • Server dashboard’s “Memory index service used”
  • Server Statistics dashboard’s “remaining index ram”
  • Indexes dashboard’s “data size”

In order to try to compute “remaining index ram” myself, I checked the response of the /api/v1/stats REST service, and had a look at the .indexer.memory_used field , which I supposed should at least include the sum of each index’ used memory (i.e. the sum of all the remaining .nameoftheindex.memory_used fields).

But I’m noticing that .indexer.memory_used considerably exceeds the sum of all the remaining .nameoftheindex.memory_used , say even by its 20%. Is this excess used by the indexer process? Or maybe I’m going astray and looking at the wrong fields …

Thank you very much in advance for any hint.

2

@giacomo.mazzini,

The <name_of_the_index>.<memory_used> stat gives the memory allocated for storing the indexed data

The indexer.memory_used field, contains the sum of all the storage allocations and also the in-memory buffers that are required to process the mutations, scans. E.g., indexer might allocated 10M of buffers to process 20M of index data. This 10M will be seen in indexer.memory_used stat

So, it is expected for indexer.memory_used to exceed sum of all <name_of_the_index>.<memory_used> stat.

Whether it should be 20% more or not depends how much your indexer memory_used it. If it is 512M, then 20% might be reasonable. If it is 50G, then we need to see the stats and understand why that is the case.

Thanks,
Varun