c# - MongoDB very slow read performance on server -
Actually, I have a Windows service that does a batch job.
My two collections are related, client accounts and events
This service runs through each account and works out of their travel times, the timestamp, the name of the event , Happened on that page, and user name event collection log actions. The risk of account closure, based on the collection of events and a set of rules defined by the user.
At this time there are around 3,500 accounts and there are about 100,000 events in my database. The service to run on my development PC only takes 1 minute, but always appears on the server (I have been guessed that It takes about 2.5 hours depending on the modification of the service, so it only works on one client account.
My machine has a core i7 with 16 GB RAM, the server has an Intel Xeon E5- 2609 (64 bit, Win 2008 R2) is with 24 GB RAM. I took the database nearly a couple of minutes to run a very old server (32 bit, Windows 2003) and the service. So, I know that it takes more than just a minute on my dev machine, and the old server hardware But it takes only 2 minutes, even though it is taking a case on a modern server.
Originally, Mongo Shell warned that NUMA was enabled on the server and performance problems To avoid it It has since been discontinued, but the performance has not been affected.
When I run on db.currentOp (); I have found that it is always someone Nature's "createIndexes" query (a indexed age was created earlier), even when I restore / restore the database to my dev machine and run the service / current over, the query "createIndexes" is not there, besides, Anything leap Is obtained.
Does anyone have any ideas / help on this mysterious performance issue? I will post the current open / mugostat / when necessary.
Quick Reply: I have re-established Mongo. No fancy configuration, just run the setup and fixed this problem.
The log file for a day filled with event logs for "making index" is 0.25GB because I was never working because Mongo was constantly indexing.
Comments
Post a Comment