Just started using the read: :secondary option.

Tonight, I tried to run a compact on one of our collections in one of our secondaries. And the reads continued to be handed over to that node and there were a lot of timeouts, of course.

After some time I decided to rs.remove that node until the compaction is done. But the connections were still trying to reach that node, so I had to restart the entire application, and it acted as expected.

After the compaction, I did added back the secondary, expecting to see it been hit again, but, it isn't, so I have to restart the app again. Isn the driver supposed to handle this ADD() and REMOVE() cases? And how about the COMPACT of a collection, when the database is marked as RECOVERING ?