I have a mongo replica set where all reads/writes must go via the primary. The majority of the writes to the primary are small UPDATES to relatively large documents (~500kB). Each document may have between 5 and 10 updates one after the other.
Recently I noticed something odd, in that the network IN to the secondaries is MUCH higher than the network IN to the primary. Given that the secondaries should only be handling replication and nothing else I'm at a bit of a loss to explain this effect.
Here is the network IN over the last hour to the PRIMARY:
And here is the network IN over the last hour for the two SECONDARIES:
The network IN on the secondaries appears to average about 200MB (per minute), which is 3 times higher than the 60MB (per minute) of the primary. What could explain this?