Wrong pool metrics when invalidating an idle resource about reactor-pool HOT 4 CLOSED

it definitely shouldn't decrease the acquired counter, and that by itself is a bug.

however, I don't know if it should be allowed to invalidate() at all in this case. if the resource is idle/has been released, it means the caller has relinquished control of the resource back to the pool. further release() and invalidate() calls should be no-op at that point, shouldn't they?

from reactor-pool.

@simonbasle let say that you use a connection for a given request (HTTP) then you return this connection to the pool, meanwhile the connection is closed. If we cannot invalidate the resource as it is not usable anymore, the resource will stay in the pool until there is an acquire operation that will remove it from the pool, however that acquire operation may happen at any time in the next seconds or the next hour. On the other hand may be we can utilise the background cleanup.

from reactor-pool.

Yeah I understand, but that's why we added background eviction.

From a glance at the codebase, the assumption that invalidate() is NO-OP after release() is there throughout the code. You shouldn't hold to a PooledRef after having called either of its release() and invalidate() method.

That is a far wider change with potential breaking implications 😞

from reactor-pool.

yeah, closing

from reactor-pool.