Comment 5 for bug 1578827

Customer has environment with 8 OSD nodes:

1) older hardware: intel oem chassis, qty 2 Intel(R) Xeon(R) CPU E5-4650 0 @ 2.70GHz. 2 chassis have 8 ssd OSD, 1 chassis has 16 ssd OSD.
newer hardware: supermicro chassis, qty 2 Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz. all chassis (5) have 23 ssd OSD.
2) switch is configured identically for all ceph nodes
3) no heavy cloud activity

ceph osd tree attached.

processes per ceph-osd node:

[root@fuel tmp]# for i in `~/bin/nodelist ceph`; do echo -ne "node-$i:\t "; date; ssh node-$i ps -elfT 2> /dev/null | wc -l; done
node-98: Thu May 5 21:59:35 UTC 2016
28930
node-99: Thu May 5 21:59:37 UTC 2016
15356
node-100: Thu May 5 21:59:38 UTC 2016
16472
node-187: Thu May 5 21:59:39 UTC 2016
33097
node-223: Thu May 5 21:59:42 UTC 2016
29964
node-224: Thu May 5 21:59:43 UTC 2016
31856
node-225: Thu May 5 21:59:45 UTC 2016
30645
node-226: Thu May 5 21:59:46 UTC 2016
30637