NodeBB Pods non response after dumping 12 lack data using jmeter scripts
-
We are trying to indentify the NodeBB benchmark for Sunbird Platform
We are expecting 1200 to 1500 tps for the below infra(specifications)
NodeBB pods => HPA(autoscaling) from min 2 to 20 pods(each pod 1 core with 1GB memory)
MongoDB => 16 cores with 32GB memoryWe noticed all userslug api's & topic related POST api's are very low performance.
We have also noticed For topic create API CPU utilisation of mongod is reaching 100% for 32 cores mongo mechine.
Find the below MongoDB load over 24 hrs
Fing the below snapshot of NodeBB load for
-
Topic create API
-
Topic read API's
With cocurrency 10 & 100
Below is the snapshot of API's throuhput & other details of Jemeter script output
Any suggestions/reason why Topic create API is taking 100% CPU usage?
-
-
@vinu said in NodeBB Pods non response after dumping 12 lack data using jmeter scripts:
Any suggestions/reason why Topic create API is taking 100% CPU usage?
Sounds like your database is the bottleneck here, not NodeBB.
As evidenced by:
We have also noticed For topic create API CPU utilisation of mongod is reaching 100% for 32 cores mongo mechine.
-
@julian Thanks for the quick response.
Any suggestions about MongoDB perforamance tunining? Especially for wirte operations?
We Just tried using both Mongo & Redis, but we have noticed redis is only used to manage user sessions.
Can we expect better performance, if we switch to Redis instead of MangoDB?
Hope out team also faced these kind of issues which doing NodeBB benchmarking. Please share if you have any references.
-
@volanar Thans for the suggestion. After switching to Postgres we are still not getting the better troughput than redis.
@julian Is it possibl to share the any benchmark results of MongoDB.
The only option we have now to go with redis(RDB + AOF) with the above results shared by @venkat123 here
https://community.nodebb.org/post/84368Any suggestions on MongoDB, Redis or Postgress config setting to get better performance.