Hi there!
I’ve a use case where I’m looking at ingesting few millions of documents to a collection, and I experience slow downs in insertions beyond a few millions.
While investigating the slow query in mongod logs, I noticed that I’m doing quite a bit of bytesRead [Mongodb Logs what does bytesRead mean in slow query log] as part of inserts, indicating cache is not used optimally.
Also observed my indexSize was around 9GBs, and I’ve wiredTiger cache of 32GB, however only 2GB index is loaded to wiredTiger cache.
There are no explicit indexes, this is the default index on the _id field.
db.getCollection("coll1").stats({indexDetails: true}).indexDetails["_id_"].cache['bytes currently in the cache']
1725358239 //1.7GB
totalIndexSize: 8959066112 //9GB
I suppose reads from disk are the root cause of the slow query, and could be improved by loading the complete index to wiredTiger cache. Could someone please help me in loading entire/larger portion of index to cache?
Thanks!