High Startup Resources

[sidhant92]

Hi, I am trying to deploy Micronaut v3.5.4 to production. What I noticed is, that when I try to directly run the jar it takes about ~700 ms to startup and around ~25% CPU when given a single core. But the same changes to ~1800ms and 300% CPU when starting the same jar with datadog-java agent.
Is this really expected because it is beating the purpose of fast startups and low resources overhead?

[sidhant92]

The used memory just after bootup is also significantly more

[mcculls]

Hi @sidhant92 the tracer does need to use CPU and heap during startup (and warmup of the application) to analyse which classes to instrument for tracing purposes. We are looking at ways to reduce this overhead and I'll update this issue when we have something you can test.

[mcculls]

The 1.14.0 release contains a number of startup improvements that could help here

[mcculls]

Startup overhead was reduced by #5026 - this feature also lets you persist type-matching data using DD_RESOLVER_CACHE_DIR=<writable-directory> which means less overhead when restarting.

There have been further minor improvements since then, I recommend trying the latest release 1.20.0

[sidhant92]

@mcculls is there any possibility of caching the entire bytecode transformations? for example once I run with the agent, it does all the transformations at runtime and from future runs I dont need to do this?

Use case: due to the agent during initial boot CPU requirement is high to bring the container up within a certain time period. Want to reduce that time to bring the container up