|
The messages are happening because we do not set up the elapsed timer until we start to listen for incoming connections, since we use select() to run the timer.
We need the timer during index builds because the record-in-RAM caching needs it. With the timer out of service, the record-in-RAM caching still works, but might be degraded if initial index builds take longer than 90 seconds. We can probably suppress these messages during initial index builds because they are benign in this case.
Not being able to connect to mongod is by design; no "waiting for incoming connections" message has appeared in the log yet, and this is the same behaviour as when a journal restore happens, or a journal file preallocation happens. No connections are allowed until these tasks are complete. (A rollback happens after we start listening, but this is perhaps behavior that ought to occur before listening starts)
|