Server hosts two mongod instances:

1. First mongod contains single 256 Mb capped collections with 100-300 inserts/sec and one index (no replica set, no sharding).
2. Second mongod is a replica set member.

It has been working fine for about year, but since day before yesterday both instances began to lot of log messages like this:

query admin.$cmd ntoreturn:1 command:

{ getlasterror: 1 }

reslen:83 1271310319ms
insert db_name.coll_name 1271310319ms

or in case of replica set

query admin.$cmd ntoreturn:1 command:

{ replSetHeartbeat: "vsrv", v: 12, pv: 1, checkEmpty: false, from: "213.xxx.xxx.xxx:27017" }

reslen:132 1271310319ms

Servers with other replica set member doesn't produce this messages.

Maybe this issue connected with the fact that kernel began to log this messages to syslog:
Apr 1 00:51:37 vsrv-1 kernel: [41610.222721] BUG: soft lockup - CPU#0 stuck for 101s! [mongod:3853]