reaper doesn't use the same tombstone timestamp
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Object Storage (swift) |
Fix Released
|
Undecided
|
Alistair Coles |
Bug Description
When the reaper issues direct_delete's it doesn't stick any timestamps in the headers - so the direct_client helpfully makes one up - a new one for each replica resulting in a number of different tombstones on disk causing a bunch of useless replication traffic.
vagrant@saio:~$ find /srv/node*
/srv/node1/
/srv/node2/
/srv/node3/
/srv/node3/
/srv/node4/
/srv/node4/
The best part of this bug is how many times people will confuse the abbreviations of timestamp and tombstone.
Changed in swift: | |
status: | New → Confirmed |
assignee: | nobody → Alistair Coles (alistair-coles) |
Changed in swift: | |
milestone: | none → 2.4.0 |
status: | Fix Committed → Fix Released |
Fix proposed to branch: master /review. openstack. org/174381
Review: https:/