I took a bit of time to get a feel for how different the straw2 mappings are vs straw1 mappings. For a bucket in which all weights are the same, I saw no changed mappings, which is as expected. However, on a map with 3 hosts each of which has 4 osds with weights 1,2,3, and 4 (crush-different-weight.straw), 1360/10000 mappings wind up different. Some good news is that this effect is smaller with smaller weight differences. When the osd weights are 1, 1.1, 1.2, and 1.3 (crush-different-weight-close.straw), only 377/10000 single osd mappings are different. I've attached a tarball with a script for reproducing the results and the maps I tested. -Sam
Attachment:
strawtest.tgz
Description: application/compressed-tar
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
