channelz related memory leak (ServerHttp2Session)

[sam-la-compass]

Problem Description
Upgrading from @grpc/grpc-js from 1.3.x to 1.5.x seems to introduce a memory leak
There seems to be an excess amount of ServerHttp2Sessions created (and unable to be freed)

We currently run a production service that has a single grpc-js server that creates multiple grpc-js clients. The clients are created and destroyed using lightning-pool.

Channelz is disabled when we initialize the server/clients as well with the 'grpc.enable_channelz': 0

Reproduction Steps
Create a single grpc-js server that calls grpc-js clients as needed from a pool resource. In our case, the server is running and when requests are made, we acquire a client via the pool (factory created once as a singleton) to make a request. These should be able to handle concurrent/multiple requests.
'grpc.enable_channelz': 0 for server/client options.

Environment

• OS Name: macOS (locally testing) and running on AWS EKS clusters (production)
• Node Version: 14.16.0
• Package Name and Version: @grpc/[email protected]

Additional Context
Checking out some heap info, it looks like there is a lot of ServerHttp2Sessions created and being retained. The attached screenshot is the service running for about an hour with limited requests (just idling). I also have another snapshot where the service is running for much longer with many more objects created.

I also read through #1958 and #1941 which seems related but had no luck getting to a solution.

Happy to provide more context or help as neeeded.

[murgatroid99]

Ironically, it looks like the bug specifically caused a memory leak when 'grpc.enable_channelz': 0 is set, because of a missing check for that in one place. I published a change that should fix it in version 1.5.10.

[sam-la-compass]

Thanks a lot for the quick fix! Upgrading to 1.5.10 looks to have fixed the issue

[sam-la-compass]

False positive, looks like there is still something lingering. Investigating to find more details, will recomment

[sam-la-compass]

ServerHttp2Session is fixed and no longer being created but the memory leak for our system still exists. Will need to reopen or create another issue once I narrow down the source