ntpq -c pe - ceph clock sync skew

(the bad node):~# ntpq -c pe
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
+dns             141.4.4.26       2 u   70  128  377    0.432   -0.898   0.106
*dns3            141.4.4.26       2 u  136  128  377    0.234    0.932   0.132
 localhost       .INIT.          16 l    -  512    0    0.000    0.000   0.000

(one of 3 good nodes):~# ntpq -c pe
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
*dns             141.4.4.26       2 u  504 1024  377    0.508   -1.887   0.206
+dns3            141.4.4.26       2 u  242 1024  377    0.266   -0.253   0.136
 localhost       .STEP.          16 l    - 1024    0    0.000    0.000   0.000

$ ceph -s
  cluster d3eb47e8-358a-41d8-af7e-92280e200a70
   health HEALTH_WARN clock skew detected on mon.node70, mon.node71, mon.node72
   monmap e1: 4 mons at {node69=10.3.20.70:6789/0,node70=10.3.20.71:6789/0,node71=10.3.20.72:6789/0,node72=10.3.20.73:6789/0}, election epoch 10, quorum 0,1,2,3 node69,node70,node71,node72
   osdmap e79: 12 osds: 11 up, 11 in
    pgmap v1600: 192 pgs: 192 active+clean; 8996 MB data, 19467 MB used, 34159 GB / 34178 GB avail
   mdsmap e1: 0/0/1 up


dns:~$ ntpq -p
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 localhost       .STEP.          16 l    - 1024    0    0.000    0.000   0.000
*time1.hzg.de    .PPS.            1 u  934 1024  377    0.830    0.524   0.669

dns3:~$ ntpq -p
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 localhost       .STEP.          16 l    - 1024    0    0.000    0.000   0.000
*time1.hzg.de    .PPS.            1 u  469 1024  377    0.795   -1.209   0.302