Apache Nifi 中用户“user”@“localhost”（使用密码：YES）的访问被拒绝

Question

I am trying to use Apache Nifi for loading tweets and storing them in memsql database. Loading and processing tweets works great, but when I come to steps of inserting to memsql, more precise when flow comes to ConvertJSONToSql processor, it throws error.

This is my setup for JDBC Connection Pool:

Memsql is running on Docker, whose cluster I initialized like this:

docker run -i --init `
    --name memsql-ciab `
    -e LICENSE_KEY=$env:LICENSE_KEY `
    -p 3308:3306 -p 8090:8080 `
    memsql/cluster-in-a-box

Port 3306 and 8080 were already taken. I also made sure that myuser has all permissions, both to host 'localhost' and '127.0.0.1', as shown in grants.

When I connect with Windows Powershell or on localhost:8090, both logins work for 'myuser' with password 'pass'. Also both logins have permissions for database 'twitterDB'. What I found weird is, in my Docker cluster I have database memsql, which in it has table users, but I can only run:

describe memsql.users;

Everything else throws error 'ERROR 1706 (HY000): Feature 'table names conflicting with metadata tables of the same name' is not supported by MemSQL.'. Also, in my JDBC Connection pool in Nifi, if I leave user and password empty, ConvertJSONToSql throws different error, which says that ''@'localhost' user doesn't have permissions for database 'twitterDB'.*.

I have been struggling for some time now with this problem, and I don't know what else to try so any help would be great.

Answer 1

Let's separate the concerns of authentication fail and insert fail. Try changing Apache Nifi to use the user root (no password) and see if the inserts succeed. Also try inserting data using sql commands from MemSQL Studio ( localhost:8090 in your setup) or from powershell. Try this with each login: root (no password) and as myuser . These tests should help us learn more.

You didn't specify the Docker setup, so assuming Docker Desktop or another local Docker, try increasing CPUs to 4 and RAM to 4 or 8 gigs. I find the developer defaults are often smaller than what works well with cluster-in-a-box .