Wednesday, November 21, 2018

Bluestore Internals

Bluestore Internals
## Bluesotre Discussions
If WAL is full what would happen? Would writes block?

It never blocks; it will always just spill over onto the next fastest
device (wal -> db -> main). Note that there is no value to a db partition
if it is on the same device as the main partition.

Would a drastic (quick) action to correct a too-small-DB-partition
(impacting performance) is to destroy the OSD and rebuild it with a
larger DB partition?
Yes

I would check your running Ceph clusters and calculate the amount of objects per OSD.
total objects / num osd * 3

For the moment though, having multiple (4)
256MB WAL buffers appears to give us the best performance despite
resulting in large memtables, so 1-2GB for the WAL is right.

A tool to gather complete Ceph cluster information
https://github.com/42on/ceph-collect

Bluesotre onode size is 24k for average object size of 2.8MB in RBD. So average object size and count per TB can be calculated.

#Reference
http://ceph-users.ceph.narkive.com/8uPMEXNz/bluestore-osd-data-wal-db

Written with StackEdit.

No comments:

Post a Comment