Hi,

ceph uses a pseudo random distribution within crush to select the target hosts. As a result, the algorithm might not be able to select three different hosts out of three hosts in the configured number of tries. The affected PGs will be shown as undersized and only list two OSDs instead of three.

The best solution is adding more host, the second best is changing crush parameters (number_of_tries etc.) until the algorithm gives better results.


Ceph clusters with few hosts often require manual tuning to work reliably.


Regards,

Burkhard

_______________________________________________
ceph-users mailing list -- ceph-users@ceph.io
To unsubscribe send an email to ceph-users-le...@ceph.io

Reply via email to