Open the journald following descriptor earlier

Following a journal log almost always requires a descriptor to be
allocated.  In cases where we're running out of descriptors, this means
we might get stuck while attempting to start following the journal, at a
point where it's too late to report it to the client and clean up
easily.  The journal reading context will cache the value once it's
allocated, so here we move the check earlier, so that we can detect a
problem when we can still report it cleanly.

Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>

daemon/logger/journald/read.go

@@ -303,14 +303,21 @@ func (s *journald) readLogs(logWatcher *logger.LogWatcher, config logger.ReadCon
 	}
 	cursor = s.drainJournal(logWatcher, config, j, "")
 	if config.Follow {
-		// Create a pipe that we can poll at the same time as the journald descriptor.
-		if C.pipe(&pipes[0]) == C.int(-1) {
-			logWatcher.Err <- fmt.Errorf("error opening journald close notification pipe")
+		// Allocate a descriptor for following the journal, if we'll
+		// need one.  Do it here so that we can report if it fails.
+		if fd := C.sd_journal_get_fd(j); fd < C.int(0) {
+			logWatcher.Err <- fmt.Errorf("error opening journald follow descriptor: %q", C.GoString(C.strerror(-fd)))
 		} else {
-			s.followJournal(logWatcher, config, j, pipes, cursor)
-			// Let followJournal handle freeing the journal context
-			// object and closing the channel.
-			following = true
+			// Create a pipe that we can poll at the same time as
+			// the journald descriptor.
+			if C.pipe(&pipes[0]) == C.int(-1) {
+				logWatcher.Err <- fmt.Errorf("error opening journald close notification pipe")
+			} else {
+				s.followJournal(logWatcher, config, j, pipes, cursor)
+				// Let followJournal handle freeing the journal context
+				// object and closing the channel.
+				following = true
+			}
 		}
 	}
 	return