Python apm agent doesn't captured spans in aws lambda functions

[sharanraj124]

Hi, I added an elastic APM agent in my lambda function transaction are captured perfectly, I don't see any database calls in my dashboard

import sys
import json
import pymysql
import logging
from elasticapm import capture_serverless



rds_host = "******"
name = "admin"
password = "***"
db_name = "***"
port = 3306

logger = logging.getLogger()
logger.setLevel(logging.INFO)

try:
    conn = pymysql.connect(host=rds_host, user=name, passwd=password, db=db_name, connect_timeout=5)
except pymysql.MySQLError as e:
    logger.error("ERROR: Unexpected error: Could not connect to MySql instance.")
    logger.error(e)
    sys.exit()

logger.info("SUCCESS: Connection to RDS mysql instance succeeded")

@capture_serverless()
def lambda_handler(event, context):

    sql = "INSERT INTO `user` (`firstName`, `lastName`,  `gender`, `birthday`,`email`, `phone`, `shippingAddress`, `billingAddress`) VALUES(%s, %s, %s, %s, %s, %s, %s, %s)"
    val = (event["firstName"], event["lastName"], event["gender"], event["birthday"], event["email"], event["phone"], event["shipping_address_id"], event["billing_address_id"])

    with conn.cursor() as cur:
        cur.execute(sql, val)
        conn.commit()

    return {
        'statusCode': 200,
        'body': json.dumps('Data has been inserted successfully')
    }

[basepi]

The pymysql instrumentation creates a proxy object on pymysql.connect().

The problem is, that we don't instrument pymysql until the handler is invoked.

We may be able to improve this -- you'll note in the comment that we're trying not to create multiple client instances. But now that we have get_client() I don't think it should matter, we can check if a client has already been created.

For now, I think you would need to do something like this (untested):

import sys
import json
import pymysql
import logging
from elasticapm import capture_serverless


rds_host = "******"
name = "admin"
password = "***"
db_name = "***"
port = 3306

logger = logging.getLogger()
logger.setLevel(logging.INFO)

conn = None


@capture_serverless()
def lambda_handler(event, context):
    global conn
    if not conn:
        try:
            conn = pymysql.connect(host=rds_host, user=name, passwd=password, db=db_name, connect_timeout=5)
        except pymysql.MySQLError as e:
            logger.error("ERROR: Unexpected error: Could not connect to MySql instance.")
            logger.error(e)
            sys.exit()

        logger.info("SUCCESS: Connection to RDS mysql instance succeeded")

    sql = "INSERT INTO `user` (`firstName`, `lastName`,  `gender`, `birthday`,`email`, `phone`, `shippingAddress`, `billingAddress`) VALUES(%s, %s, %s, %s, %s, %s, %s, %s)"
    val = (event["firstName"], event["lastName"], event["gender"], event["birthday"], event["email"], event["phone"], event["shipping_address_id"], event["billing_address_id"])

    with conn.cursor() as cur:
        cur.execute(sql, val)
        conn.commit()

    return {
        'statusCode': 200,
        'body': json.dumps('Data has been inserted successfully')
    }

Even if we fixed the issue where we don't instrument until the function is called, you would still need to move your connection code below the @capture_serverless() call. But for now it needs to be inside the function.

[basepi]

Actually I just remembered why we can't move it out of the decorator: We allow config via the **kwargs to the decorator. That may be something that's not super important to support, but that's the reason it's set up the way it is.

I will note, we're currently working on a no-code-changes instrumentation system, which will eventually include support for lambda. This will mean you will just add our lambda layer (which will include both the python agent and the extension) and environment variables and everything should just work. We don't have an ETA on that feature, but it should solve your use case much more cleanly.