MongoDB tips: Performance Tuning, Index creation order matters
I am hoping this post will be obsolete and noone will needs to find this. Continue from my previous post about choosing the best index for the job, that certain should still holds true as when I created the index with 1, -1 for two different date field, it's less efficient than -1, -1.
Anyway, so luckily I found out about this before rolling this into production as I was planning on the production release. Under lower environment, I realized a less efficient index was used... from my previous post. I now believe it's a bug that should be fixed, in my experiments, the newer index is always preferred even when a better index is present. MongoDB query optimizer looks to be looking into only the first selective field I chose and no further. So if the latest index created matches the field this index is created, so in order for my query to serve the high frequency query, I have to make sure the best index is created last.
Anyway, so luckily I found out about this before rolling this into production as I was planning on the production release. Under lower environment, I realized a less efficient index was used... from my previous post. I now believe it's a bug that should be fixed, in my experiments, the newer index is always preferred even when a better index is present. MongoDB query optimizer looks to be looking into only the first selective field I chose and no further. So if the latest index created matches the field this index is created, so in order for my query to serve the high frequency query, I have to make sure the best index is created last.
Comments
Post a Comment