some mapping failed

Running Ceph 12.2.5, crush 1.0.35 (current PyPI version).

default ceph report does not have information about pg,so I using

```
ceph pg dump -f json > pg_dump.json
# edit crush/ceph/__init__.py,add line 72
report['pgmap'] = json.load(open('pg_dump.json'))
# replace crush_ruleset with crush_rule in line 115
ruleset = pool['crush_rule']
```

and then output is:

```
2018-07-16 16:35:04,999 42.7e map to [32, 320, 144] instead of [32, 320, 167]
2018-07-16 16:35:04,999 42.73 map to [134, 295, 144] instead of [134, 295, 196]
2018-07-16 16:35:05,000 43.57 map to [21, 416, 144] instead of [21, 416, 185]
2018-07-16 16:35:05,004 44.87 map to [41, 144, 314] instead of [41, 314]
2018-07-16 16:35:05,005 44.7d map to [144, 371, 34] instead of [227, 371, 34]
2018-07-16 16:35:05,006 45.f2 map to [144, 414, 48] instead of [237, 414, 48]
2018-07-16 16:35:05,007 45.46 map to [144, 336, 14] instead of [234, 336, 14]
2018-07-16 16:35:05,008 46.45 map to [361, 144, 47] instead of [361, 256, 47]
2018-07-16 16:35:05,008 46.65 map to [144, 377, 23] instead of [218, 377, 23]
```

Comienza con crush/ceph/__init__.py alrededor de las líneas 72 y 115, y luego reproduce el informe usando Ceph 12.2.5 y los datos de pg_dump.json de `ceph pg dump -f json`. Compara las asignaciones indicadas por el issue con las asignaciones esperadas y determina qué comportamiento debería corregirse.

Escrito por el modelo de indexación a partir del texto del issue.