Ceph

2014-04-18 17:01:34.565301 7f69feae6700 10 osd.1 pg_epoch: 77 pg[3.2s0( v 77'466 lc 40'64 (0'0,77'466] local-les=73 n=465 ec=8 les/c 73/26 72/72/72) [1,5,0] r=0 lpr=72 pi=25-71/1 luod=77'465 rops=15 crt=71'434 mlcod 38'58 active+recovering m=357] continue_recovery_op: READING return RecoveryOp(hoid=d362481a/plana1323219-759/head//3 v=40'71 missing_on=1(0) missing_on_shards=^@ recovery_info=ObjectRecoveryInfo(d362481a/plana1323219-759/head//3@40'71, copy_subset: [], clone_subset: {}) recovery_progress=ObjectRecoveryProgress(!first, data_recovered_to:2920448, data_complete:true, omap_recovered_to:, omap_complete:true) pending_read=0 obc refcount=1 state=WRITING waiting_on_pushes=1(0) extent_requested=2105344,1052672

Notice the missing_on_shards=^@ in the log message.

The ECBackend::RecoveryOp has "set<shard_id_t> missing_on_shards" that is output using operator<<() with the following:

 << " missing_on_shards=" << rhs.missing_on_shards

Since shard_id_t is a uint8_t which is really an unsigned char a value of 0 is getting logged as ^@ instead of "0".