We have two mongodb shards 29GB each. All data is about 50GB.
When we restarted our servers appeared the following behaviour:
res memory on each shard grows very slow (about 1MB/sec), but userd memory (by "free" linux command) grows very fast.
Now we looks at this situation:

mongostat:
insert query update delete getmore command flushes mapped vsize res faults locked % idx miss % qr|qw ar|aw netIn netOut conn time
15 109 9 2 57 87 0 33.9g 72.2g 1.49g 0 3.1 0 0|0 0|0 39k 329k 81 00:33:51

but:

root@MongoShard1:~# free
total used free shared buffers cached
Mem: 29367112 29226476 140636 0 40732 28061776
-/+ buffers/cache: 1123968 28243144
Swap: 1048568 0 1048568

about 29gb used, but by mongo only 1.49gb
there are no any other processes on server that can use more than some megabytes of ram

Does it mongodb bug?