Elasticsearch APM Configuration for Tomcat

[vishakh]

Kibana version: 7.8.0
Elasticsearch version: 7.8.0
APM Server version: 7.8.0
OS Platform: Amazon Linux
Tomcat Version:
Server version: Apache Tomcat/8.5.57
Server built: Jul 16 2020 22:38:29 UTC
Server number: 8.5.57.0
OS Name: Linux
OS Version: 4.14.114-83.126.amzn1.x86_64
Architecture: amd64
JVM Version: 1.8.0_252-b09
JVM Vendor: Oracle Corporation

As our deployment doesn't match the generic setup, I'm referring to Apache Tomcat- Unix instructions from: APM-Agent JAVA Configuration

Question-1: To the provided, 4 liner Catalina_Opts can I add export CATALINA_OPTS="$CATALINA_OPTS -Delastic.apm.secret_token= " parameter as-well since we need to pass the secret-key for agent

Question-2: Currently, we do have an another instrumentation tool running on the same host-server binded to same tomcat. Does having ELK-APM on the same host and binding to same Tomcat process causes any performance bottleneck/impact?

[Sylvain_Juge]

Hi @vishakh,

1. Yes, you need to add secret_token parameter, or use the newer api_key option which uses API keys, which is a bit more flexible (you can have more than one key).

2. It depends.

Performance-wise our agent is engineered to have minimal impact, and having more than one instrumentation should not change that.

However, it's likely that there may be some conflicts, for example if both agents instrument the same classes, one agent could definitely overwrite changes made by the other, which in our case would mean our agent won't be able to monitor some transactions/spans. For example, having two APM agents will very likely make them instrument the same classes, however if the purpose of the agents is not the same (for example scanning dependencies as runtime), there should not have any issue.

Also, on a more technical level, if both agents rely on bytebuddy for bytecode modification, it is more likely to make them compatible because this library is designed for it.

What is the other agent name/purpose ?