Actually, I think it's hard for opentracing-spring-cloud-jdbc-starter to fix this issue.
opentracing-spring-cloud-jdbc-starter should follow the standard of JDBC, this behavior implemented by mysql-connector-java.

I resolved this issue by add this argument to mysql jdbc url:

jdbc:mysql://localhost:3306/test?useHostsInPrivileges=false

After changing useHostsInPrivileges to false, mysql-connector-java can get the username from memory directly.

    /**
     * What's our user name as known to the database?
     * 
     * @return our database user name
     * @throws SQLException
     */
    public String getUserName() throws SQLException {
        if (this.conn.getUseHostsInPrivileges()) {
            Statement stmt = null;
            ResultSet rs = null;

            try {
                stmt = this.conn.getMetadataSafeStatement();

                rs = stmt.executeQuery("SELECT USER()");
                rs.next();

                return rs.getString(1);
            } finally {
                if (rs != null) {
                    try {
                        rs.close();
                    } catch (Exception ex) {
                        AssertionFailedException.shouldNotHappen(ex);
                    }

                    rs = null;
                }

                if (stmt != null) {
                    try {
                        stmt.close();
                    } catch (Exception ex) {
                        AssertionFailedException.shouldNotHappen(ex);
                    }

                    stmt = null;
                }
            }
        }

        return this.conn.getUser();
    }

I think the userName is not necessary at all,because normally application accesses the database by a fixed user. And then we can replace the userName by database ip and port because this is helpful to draw topology.

so the opentracing-jdbc project should support more info about database such as database ip and port

In design level, I think opentracing-jdbc is not flexiable beacuse of it's construction method.So i suggest we need to improve the opentracing-jdbc firstly and then add database ip and port to acted as tag k-v.

@jpkrohling

Now,TracingConnnection receieve a ConnectionInfo Object containing all the database information(instance,ip,port,type,username).So username is not required.And then we can reslove the database information by parse the jdbc url.For example the skywalking is doing so.

https://github.com/apache/incubator-skywalking/blob/master/apm-sniffer/apm-sdk-plugin/jdbc-commons/src/main/java/org/apache/skywalking/apm/plugin/jdbc/connectionurl/parser/ConnectionURLParser.java

@dozer47528 why close this issue,i think we need a pr to solve this problem

@pavolloffay what do you think about my suggestion?

@wuyupengwoaini would you like to contribute a fix with your solution?

Ok, i'll try to fix it within a few days

@pavolloffay please help me to review the pr