ext: fix issues in net gc task

[fantix]

• don't run on dropping databases
• don't call asyncio.wait([])
• fix wrong logging without decoding JSON

Refs #7888

[scotttrinh]

🙏

[elprans]

Why aren't we using a TaskGroup here?

[fantix]

Yeah, I guess we can; this was a carried-over from an old commit, let me try again so that we know for sure.

[fantix]

Ah it's just unnecessary, because the Task we're in here is safe without cancellation - the server doesn't need to wait for gc() itself to clean up anything, while all child Tasks are already managed by the TaskGroup in each Tenant (by tenant.create_task()).

We would need an extra coroutine wrapper to use TaskGroup here:

async def _await_task(task: asyncio.Task):
    return await task


async def gc(server: edbserver.BaseServer) -> None:
    while True:
        try:
            async with asyncio.TaskGroup() as g:
                for tenant in server.iter_tenants():
                    if tenant.accept_new_tasks:
                        task = tenant.create_task(
                            _gc(tenant, NET_HTTP_REQUEST_TTL),
                            interruptable=False,
                        )
                        g.create_task(_await_task(task))
...

[msullivan]

... looks like this doesn't fix that DROP flake though