With latest version of druid is it possible to use OpenStack Swift custom implementation (not Amazon S3)?

[anantmf]

With latest version of druid is it possible to use OpenStack Swift custom implementation (not Amazon S3) as a deep storage?
I am having some issue in setting up this, so wanted to check if this is supported or not before asking further questions.

Just FYI, following properties I am setting

druid.s3.endpoint.url=custom url
druid.storage.type=s3
druid.storage.bucket=bucket
druid.storage.baseKey=druid/segments
druid.s3.accessKey=actual key
druid.s3.secretKey=key

[jihoonson]

Hi @anantmf, it should work if you use S3 compatible API for Swift.

[jihoonson]

If you want to use http, you may need #6954.

[anantmf]

Hi @jihoonson , Thanks for the reply. I found couple of configuration issues after which I was able to make it work. So now segments are correctly stored in Swift, however when Historical is trying to get it back from Deep storage it fails.

I think the issue is
#5382 (comment)
This is exactly what is happening in my case. Where result.getKeyCount() is 0 but result.getObjectSummaries() have size > 0.

When it is doing ListObjectsV2Request, i tried to debug the code and the response actually has object summary but KeyCount is Zero. Which is causing it to fail.
I tried to debug the code of unmarshalling of List object result, and it seems the Unmarshaller expects the response to have a element called as KeyCount then only it sets it in the result object, otherwise it stays zero.

This is the actual log I get:
2019-02-02T15:04:19,789 ERROR [ZkCoordinator] org.apache.druid.segment.loading.SegmentLoaderLocalCacheManager - Failed to load segment in current location /data/projects/imply-2.8.9/var/druid/segment-cache, try next location if any: {class=org.apache.druid.segment.loading.SegmentLoaderLocalCacheManager, exceptionType=class org.apache.druid.segment.loading.SegmentLoadingException, exceptionMessage=Cannot find object for bucket[test4] and key[druid/segments/Consumer_Complaints/2016-01-01T00:00:00.000Z_2017-01-01T00:00:00.000Z/2019-02-02T15:03:18.693Z/0/index.zip], location=/data/projects/imply-2.8.9/var/druid/segment-cache}
org.apache.druid.segment.loading.SegmentLoadingException: Cannot find object for bucket[test4] and key[druid/segments/Consumer_Complaints/2016-01-01T00:00:00.000Z_2017-01-01T00:00:00.000Z/2019-02-02T15:03:18.693Z/0/index.zip]
at org.apache.druid.storage.s3.S3DataSegmentPuller.getSegmentFiles(S3DataSegmentPuller.java:139) ~[?:?]
at org.apache.druid.storage.s3.S3LoadSpec.loadSegment(S3LoadSpec.java:60) ~[?:?]
at org.apache.druid.segment.loading.SegmentLoaderLocalCacheManager.loadInLocation(SegmentLoaderLocalCacheManager.java:200) ~[druid-server-0.13.0-incubating-iap6-SNAPSHOT.jar:0.13.0-incubating-iap6-SNAPSHOT]
at org.apache.druid.segment.loading.SegmentLoaderLocalCacheManager.loadInLocationWithStartMarker(SegmentLoaderLocalCacheManager.java:188) ~[druid-server-0.13.0-incubating-iap6-SNAPSHOT.jar:0.13.0-incubating-iap6-SNAPSHOT]
at org.apache.druid.segment.loading.SegmentLoaderLocalCacheManager.loadSegmentWithRetry(SegmentLoaderLocalCacheManager.java:151) [

[jihoonson]

Thank you for the investigation! I haven’t expected that such case would happen. Would you please raise a PR to fix it? I think we can check the size of objectSummaries instead of keyCount.

[anantmf]

@jihoonson Raised a PR #7000
This is my first OpenSource PR, so let me know if anything I have missed.

[MrMoronIV]

May I ask what you've used for connecting Druid to Swift? I can't find anything about it in the documentation.

Is it supposed to work using the S3 compatible API?

[stale]

This issue has been marked as stale due to 280 days of inactivity. It will be closed in 4 weeks if no further activity occurs. If this issue is still relevant, please simply write any comment. Even if closed, you can still revive the issue at any time or discuss it on the dev@druid.apache.org list. Thank you for your contributions.

[stale]

This issue has been closed due to lack of activity. If you think that is incorrect, or the issue requires additional review, you can revive the issue at any time.