Sporadic java.lang.OutOfMemoryError: GC overhead limit exceeded

[#1]

Hello!

I'm still evaluating a free trial account and have a moderately complex API Proxie. After adding a few more policies to the flow, I started getting the following error ~50% of the time:

error       GC overhead limit exceeded
type        ErrorPoint
state       PROXY_REQ_FLOW
error.class java.lang.OutOfMemoryError
Identifier  fault

I suspect that this is from a limitation of the free trial account. When I upgrade to SMB, will this error disappear?

Thank you.

[DChiesa]

It's hard to say!

For example, I can imagine building an API proxy with a Java callout, and the logic in the Java callout simply allocates memory in a loop, until it fails. This will cause an out-of-memory error in any version of Apigee Edge.

To give you a better answer, we'd need to know more context - just what is your API Proxy doing, are you using custom java code, and under which conditions do you see this error? Are you sending very large payloads, and are you sending many concurrent requests?

Aside from all of that, suspecting the free trial account seems prudent. There are numerous tenants that share every single JVM on the free trial service. We try to make it as reliable and as resilient as possible, and in fact we track down such "out of memory" errors as diligently as we can. But sometimes there are problems due to "greedy neighbors".

[#3]

I'm not spinning up any loops. My request payloads are small (<1k).

The GC error happens on the the very first policy hit (Verify API Key in the proxy pre-flow).

My proxy uses 28 policies with three target endpoints. They are mostly extract/assign vars. I have one Service Callout and a couple Javascript scripts. But again, the error doesn't let me past the first policy.

I can easily believe that I'm exceeding the free tier "test things out" limits. Do you have an estimate of how much more memory is allocated for SMB accounts? 10x? 100x? More?

[#4]

Please give me your org name. I will have check on this issue.

[#5]

The org is shoutpoint.

I just did a little testing and have not experienced the out-of-memory error again. I last saw the error about 22 hours ago.

I don't plan on working in/testing Apigee for the next few days. Still, I'll post when the error happens again. Are there additional details of this error that would help? A trace log? Timestamp?