Ceph » rgw

There is such a scenario. After executing the "InitiateMultipartUpload" interface, we execute the "UploadPart" interface to upload data. When this action is not completely executed, for some reason, we call "AbortMultipartUpload" to terminate the upload, and finally we pass "radosgw" -admin bucket list" command to view the files in this bucket, it will display the shard left by "UploadPart", that is, the shard has not been deleted.(as shown in 1.jpg)
Going further, we verified whether there is complete data in the pool, but found that there is no sharded data in the data pool, but there is index data in the index pool.(as shown in 2.jpg)

We went to find the reason and found that because we used "AbortMultipartUpload" to delete the part information, an error was reported during the execution of the "void RGWPutObj::execute()" function, but at the end of this function, the data in the data pool was deleted, and The index data in the index pool is not deleted.

So what we are confused about is why this phenomenon occurs and how to solve it.