Apparently we don't strip EC sysmeta when copying data between policies. I'm *pretty sure* if you're going between two different EC policies, it'll be fine (surely we would've heard about that bug by now otherwise, right?). Going from EC to replicated still *works* (since the replicated data path won't look at the headers), but it leads to some head-scratchers when debugging:
vagrant@saio:~/swift$ swift post c -H x-storage-policy:ec
vagrant@saio:~/swift$ swift upload c tox.ini
tox.ini
vagrant@saio:~/swift$ swift post c2 -H x-storage-policy:replicated
vagrant@saio:~/swift$ swift copy c tox.ini --destination /c2
created container c2
c/tox.ini copied to /c2/tox.ini
vagrant@saio:~/swift$ swift-get-nodes /etc/swift/object-1.ring.gz AUTH_test/c2/tox.ini | grep curl
curl -g -I -XHEAD "http://127.0.0.3:6036/sdb3/236/AUTH_test/c2/tox.ini" -H "X-Backend-Storage-Policy-Index: 1"
curl -g -I -XHEAD "http://127.0.0.1:6016/sdb1/236/AUTH_test/c2/tox.ini" -H "X-Backend-Storage-Policy-Index: 1"
curl -g -I -XHEAD "http://127.0.0.4:6046/sdb4/236/AUTH_test/c2/tox.ini" -H "X-Backend-Storage-Policy-Index: 1"
curl -g -I -XHEAD "http://127.0.0.2:6026/sdb2/236/AUTH_test/c2/tox.ini" -H "X-Backend-Storage-Policy-Index: 1" # [Handoff]
vagrant@saio:~/swift$ curl -g -I -XHEAD "http://127.0.0.3:6036/sdb3/236/AUTH_test/c2/tox.ini" -H "X-Backend-Storage-Policy-Index: 1"
HTTP/1.1 200 OK
Content-Type: application/octet-stream
X-Object-Sysmeta-Container-Update-Override-Etag: 0XqAb6vL1glzuefZ9COiEjByMg2Z2IVYNgHCMLV4+5c=; swift_meta=%7B%22cipher%22%3A+%22AES_CTR_256%22%2C+%22iv%22%3A+%22KsLvGHOR4YmK89uKQQ785A%3D%3D%22%2C+%22key_id%22%3A+%7B%22path%22%3A+%22%2FAUTH_test%2Fc2%2Ftox.ini%22%2C+%22secret_id%22%3A+%222018%22%2C+%22v%22%3A+%222%22%7D%7D
X-Object-Sysmeta-Ec-Etag: 1d2d2a97ac44b3b0a26b5379f427af69
X-Object-Sysmeta-Ec-Content-Length: 4788
X-Object-Sysmeta-Ec-Frag-Index: 4
X-Object-Sysmeta-Ec-Scheme: liberasurecode_rs_vand 4+2
X-Object-Sysmeta-Ec-Segment-Size: 1048576
X-Object-Transient-Sysmeta-Crypto-Meta-Mtime: St6CSFvUClCuZRBBld5Ynik=; swift_meta=%7B%22cipher%22%3A+%22AES_CTR_256%22%2C+%22iv%22%3A+%226SObRXmr8tWSSF6%2FobRFxg%3D%3D%22%7D
X-Object-Transient-Sysmeta-Crypto-Meta: %7B%22cipher%22%3A+%22AES_CTR_256%22%2C+%22key_id%22%3A+%7B%22path%22%3A+%22%2FAUTH_test%2Fc2%2Ftox.ini%22%2C+%22secret_id%22%3A+%222018%22%2C+%22v%22%3A+%222%22%7D%7D
X-Object-Sysmeta-Crypto-Etag: FgnZOAjxo2Rb0DIBbVeEX25LcrspMDW0/WHBk8NqKFM=; swift_meta=%7B%22cipher%22%3A+%22AES_CTR_256%22%2C+%22iv%22%3A+%22q02K12yOKR1tAVkitnD4pQ%3D%3D%22%7D
X-Object-Sysmeta-Crypto-Body-Meta: %7B%22body_key%22%3A+%7B%22iv%22%3A+%22eLyfKWyPM7ghVpTrBXtJJw%3D%3D%22%2C+%22key%22%3A+%22L8Vb1%2B2w4GjEXljd3kvitUOkZBMJt9%2BTQIWFlJdSN5s%3D%22%7D%2C+%22cipher%22%3A+%22AES_CTR_256%22%2C+%22iv%22%3A+%22KTHSdioYOinQTzoaMkIpDw%3D%3D%22%2C+%22key_id%22%3A+%7B%22path%22%3A+%22%2FAUTH_test%2Fc2%2Ftox.ini%22%2C+%22secret_id%22%3A+%222018%22%2C+%22v%22%3A+%222%22%7D%7D
X-Object-Sysmeta-Crypto-Etag-Mac: j7a0gL63t8MxcSlBx1+GJ+7O1zmxwiLFCQHW1fqsJLU=
Etag: "236f1845de751090ce30b740081a5c4b"
Last-Modified: Wed, 01 Jul 2020 04:57:02 GMT
X-Timestamp: 1593579421.76668
X-Backend-Timestamp: 1593579421.76668
X-Backend-Data-Timestamp: 1593579421.76668
X-Backend-Durable-Timestamp: 1593579421.76668
Content-Length: 4788
Date: Wed, 01 Jul 2020 04:57:56 GMT
Note how Content-Length and X-Object-Sysmeta-Ec-Content-Length match *exactly* -- that should basically *never* happen for real EC data.
Reviewed: https:/
Committed: https:/
Submitter: "Zuul (22348)"
Branch: master
commit 399bda8a410893b
Author: Tim Burke <email address hidden>
Date: Tue Mar 30 15:25:13 2021 -0700
ec: Don't copy EC metadata to replicated objects
Change-Id: Ia6c1b75a24d2b4
Closes-Bug: #1885857