Absent "database" attribute causes a defect when integrating with opentelemetry-instrumentation-dbapi

[cazorla19]

Describe the bug

We're currently implementing an OpenTelemetry instrumentation to our application which uses mysqlclient using opentelemetry-instrumentation-dbapi module. While it generally works, it's missing the DB name in our database spans generated by this module:

{
    "name": "INSERT",
    "context": {
        "trace_id": "0x5850dafa35e0e4863d252eca319f4a52",
        "span_id": "0xe9e9bb17d01d11aa",
        "trace_state": "[]"
    },
    "kind": "SpanKind.CLIENT",
    "parent_id": "0x995b0ebffa48c274",
    "start_time": "2024-12-16T10:54:37.238992Z",
    "end_time": "2024-12-16T10:54:37.240130Z",
    "status": {
        "status_code": "UNSET"
    },
    "attributes": {
        "db.system": "mysql",
        "db.name": "", # it creates the attribute, but it's empty, because it can't get the database name
        "db.statement": "INSERT INTO sample_table (name, age) VALUES (%s, %s)",
        "net.peer.port": 3306
    },
    "events": [],
    "links": [],
    "resource": {
        "attributes": {
            "telemetry.sdk.language": "python",
            "telemetry.sdk.name": "opentelemetry",
            "telemetry.sdk.version": "1.29.0",
            "service.name": "test_app",
            "db.name": "test_db" # here it gets passed
        },
        "schema_url": ""
    }
}

I've had a look at the instrumentation module's code to find out why it fails to get the database name.

https://github.com/open-telemetry/opentelemetry-python-contrib/blob/main/instrumentation/opentelemetry-instrumentation-dbapi/src/opentelemetry/instrumentation/dbapi/__init__.py#L376

self.database = self.connection_props.get("database", "")

How the module finds the DB name is that it looks up for the self.database attribute of the Connection object. You can redefine it to any other attribute where the DB name is stored (like db attribute, for example), but the main thing is that there should be one attribute in Connection object where the DB name is stored.

The mysqlclient is missing it, while other Python modules have such an attribute:

In [26]: import pymysql
    ...: 
    ...: conn = pymysql.connect(host="localhost", user="root", db="test_db")
    ...: print(conn.db.decode())
test_db

In [27]: import mysql.connector
    ...: 
    ...: conn = mysql.connector.connect(host="localhost", user="root", database="test_db")
    ...: print(conn.database)
test_db

It seems to be a minor/non-critical issue, but here we're losing an opportunity to select/filter spans by the database name, which makes the troubleshooting and observability experience better.

As it's basically one line of code assigning the database name from connection parameters to the object attribute, I decided to do add one and propose the change. Please take a look at my pull request #752

Environment

this issue is environment-independent

How did you install libmysqlclient libraries?

No response

What version of mysqlclient do you use?

No response

Docker command to start MySQL server

No response

Minimum but complete code to reproduce

import MySQLdb
from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter
from opentelemetry.trace import set_tracer_provider
from opentelemetry.instrumentation.dbapi import trace_integration

DB_NAME = "test_db"

# Set up the Console exporter
console_exporter = ConsoleSpanExporter()

# Set up OpenTelemetry tracer provider
provider = TracerProvider()
set_tracer_provider(provider)

# Add Console Exporter
provider.add_span_processor(SimpleSpanProcessor(console_exporter))

# Instrument MySQLdb
trace_integration(
    connect_module=MySQLdb,
    connect_method_name="connect",
    database_system="mysql"
)

# Get tracer instance
tracer = trace.get_tracer(__name__)

def main():
    try:
        with tracer.start_as_current_span("connect-to-database") as span:
            # Connect to the MySQL database
            connection = MySQLdb.connect(
                host="127.0.0.1",  # Use TCP to connect
                port=3306,
                user="root",
                db=DB_NAME
            )
            span.add_event("Successfully connected to the database")
            print("Connected to MySQL database")

        with tracer.start_as_current_span("perform-database-operations") as span:
            cursor = connection.cursor()

            # Execute a query to check the connection
            cursor.execute("SELECT DATABASE();")
            record = cursor.fetchone()
            print(f"You're connected to the database: {record[0]}")

            # Insert a sample record
            insert_query = "INSERT INTO sample_table (name, age) VALUES (%s, %s)"
            cursor.execute(insert_query, ("Alice", 30))
            connection.commit()
            print("Inserted record into the database.")

            # Query the data
            cursor.execute("SELECT * FROM sample_table;")
            rows = cursor.fetchall()
            print("Data from the table:")
            for row in rows:
                print(row)

    except MySQLdb.Error as e:
        print(f"Error: {e}")
    finally:
        if "connection" in locals() and connection:
            connection.close()
            print("MySQL connection is closed.")

if __name__ == "__main__":
    main()

Schema and initial data required to reproduce.

no initial data required to reproduce

Commands, and any other step required to reproduce your issue.

In [25]: import MySQLdb
...: ...: conn = MySQLdb.connect(host="localhost", user="root", db="test_db")
...: ...: print(dir(conn))
['DataError', 'DatabaseError', 'Error', 'IntegrityError', 'InterfaceError', 'InternalError', 'NotSupportedError', 'OperationalError', 'ProgrammingError', 'Warning', 'class', 'delattr', 'dict', 'dir', 'doc', 'enter', 'eq', 'exit', 'format', 'ge', 'getattribute', 'gt', 'hash', 'init', 'init_subclass', 'le', 'lt', 'module', 'ne', 'new', 'reduce', 'reduce_ex', 'repr', 'setattr', 'sizeof', 'str', 'subclasshook', 'weakref', '_binary_prefix', '_bytes_literal', '_get_native_connection', '_server_version', '_transactional', '_tuple_literal', 'affected_rows', 'autocommit', 'begin', 'change_user', 'character_set_name', 'client_flag', 'close', 'commit', 'converter', 'cursor', 'cursorclass', 'default_cursor', 'discard_result', 'dump_debug_info', 'encoders', 'encoding', 'errno', 'error', 'escape', 'escape_string', 'field_count', 'fileno', 'get_autocommit', 'get_character_set_info', 'get_host_info', 'get_proto_info', 'get_server_info', 'info', 'insert_id', 'kill', 'literal', 'messages', 'next_result', 'open', 'ping', 'port', 'query', 'read_query_result', 'rollback', 'select_db', 'send_query', 'server_capabilities', 'set_character_set', 'set_server_option', 'set_sql_mode', 'show_warnings', 'shutdown', 'sqlstate', 'stat', 'store_result', 'string_literal', 'thread_id', 'use_result', 'warning_count']

[cazorla19]

closing this one, as #752 got merged

[methane]

#752 didn't fix it. See #753. It supports db.name, db.user, and net.peer.name.