pg stuck in activating because of dropped pg-temp message
1. mon sent pg-create(1.0) to osd.1
2. osd.1 created 1.0 with acting set of [1,0]. pg 1.0 was active+clean by then.
3. osd.0 was down
4. osd.1 consumed the new osdmap, the up and acting set of pg 1.0 were [1,2], it started peering.
5. osd.1 notified osd.2 that it should create replica of pg 1.0
6. osd.2 withheld the creating of pg 1.0 because its "max-pg-per-osd" was 1
7. osd.2 removed all created pgs
8. osd.2 tried to a osd-pgtemp to mon. but the connection was reset.
9. osd.2 reconnected to mon, but the osd-pgtemp was not resent because the osd-mon connection is lossy.
10.osd.2 never resent the osd-pgtemp..