Thank you Igor, Yeah the 25K waste per rados object seems reasonable, couple of questions though: 1. Is the story of blobs re-using empty sub-sections of already allocated "min_alloc_size"ed blocks, just for RBD/FS? I read some blogs about onode->extent->blob->min_alloc->pextent->disk flow and how a write smaller than min_alloc_size counts as a small write and if any other small write comes by later, which is fit to this empty area of the blob, it will use that area; I expected this re-use behavior in rados overally. Is my assumption of how allocation works totally wrong or it just doesn't apply for s3? (maybe because object hints are immutable?) and do you know any documentation about allocation details? I couldn't find much official data about it. 2. We have a ceph cluster that was updated to pacific but the OSDs were from a previous octopus cluster with bluestore_min_alloc_size_hdd = 64KB, bluestore_prefer_deferred_size_hdd = 65536, and both bluestore_allocator and bluefs_allocator are bitmap and OSDs were not re-deployed afterward. We were concerned that re-deploying HDD OSDs with bluestore_min_alloc_size_hdd = 4KB might cause i/o performance issues since the number of blocks and hence write operations will increase. Do you know how it might affect the cluster? Many thanks _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |