[Re-open in 2015 Oct] DELETE operation not write affinity aware
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Object Storage (swift) |
Fix Released
|
Medium
|
Lingxian Kong |
Bug Description
This suppose to be fixed in an old report :
DELETE operation not write affinity aware
https:/
I found this issue appears in recent swift release. My test environment is a 3 more regions for 3 replica policy.
User performs DELETE before the object been replicated to other regions.
```
HugotekiMacBook
Error Deleting: c4/1: Object DELETE failed: http://
```
Here's a screenshot of my operation : http://
Seems the request was not been past into best_response in this case.
R1Node1 - Oct 6 07:16:52 ip6-localhost object-server: 192.168.99.1 - - [06/Oct/
R2Node2 - Oct 6 07:16:52 ip6-localhost object-server: 192.168.99.1 - - [06/Oct/
R3Node3 - Oct 6 07:16:52 ip6-localhost object-server: 192.168.99.1 - - [06/Oct/
Proxy - Oct 6 07:16:52 ip6-localhost proxy-server: 60.251.42.102 60.251.42.102 06/Oct/
Changed in swift: | |
status: | New → Confirmed |
importance: | Undecided → Medium |
Changed in swift: | |
status: | Confirmed → In Progress |
Seems like the previous patch would only map 404->204 on deletes if there was no quorum. But if there was a quorum of 404s, it stiff returns a 404.
In the provided example, the object servers returned 404, so the proxy also returned 404. But note that the tombstones are written to disk.
This might be a bug with the CLI reporting an error when there isn't one.