The handoff partition `24228` keeps on the disk. The reconstructor doesn't remove it even the primary location has this fragment. Perhaps it's garbage fragment from old swift version?
```
[root@swift-node15 objects-1]# pwd
/srv/node/d329/objects-1
[root@swift-node15 objects-1]# ls 24228/3d3/5ea48f8dbc06abf1369634fee53d13d3/1528179813.50318#0.data
24228/3d3/5ea48f8dbc06abf1369634fee53d13d3/1528179813.50318#0.data
[root@swift-node15 objects-1]# /opt/ss/bin/python /opt/ss/bin/swift-object-reconstructor ~/ss-agent/2.conf -o -v -d d329 -p 24228
object-reconstructor: Starting 37988
object-reconstructor: Spawned worker 38189 with {'override_devices': ['d329'], 'override_partitions': [24228]}
object-reconstructor: Running object reconstructor in script mode.
object-reconstructor: Run listdir on /srv/node/d329/objects-1/24228
object-reconstructor: 1/15954 (0.01%) partitions reconstructed in 0.17s (5.80/sec, 45m remaining)
object-reconstructor: Object reconstruction complete (once). (0.00 minutes)
object-reconstructor: Forked worker 38189 finished
object-reconstructor: Worker 38189 exited
object-reconstructor: Finished 37988
object-reconstructor: Exited 37988
```
```
Swift Package Version: 2.16.0.2-1.el7
Operating System Version: Linux-3.10.0-514.16.1.el7.x86_64-x86_64-with-centos-7.3.1611-Core
```
Some maybe-related bugs:
- https:/
- https:/
So is there *any* HTTP traffic with the other nodes? Is the fragment durable on the other nodes? What's the reclaim_age like?