[
https://issues.apache.org/jira/browse/HBASE-20431?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16450171#comment-16450171
]
Andrew Purtell commented on HBASE-20431:
----------------------------------------
Thanks for commenting [[email protected]].
For others: S3Guard is HADOOP-13345. Requires Hadoop 2.9.0, which we wouldn't
do for the general case, but if leaning heavily on S3A (and I'd like to do that
if we can), then this would be fine, given the dependency.
> PUT-COPY is atomic
Normally we are dealing with only a single file to move, so this is good to
have confirmation of. We also have in our design scope running against Ceph's
radosgw so I don't know if we can rely on it totally, but we can take advantage
of it if we detect we are running against S3 proper.
By the way in testing I noticed some times we'd get back (paraphrased) "200
Internal Error, please retry" from PUT-COPY, which is alarming because the
result code of 200 for an internal error is nonsense, but the embedded
AmazonClient in S3A was able to detect the error all the same and throw. But
this makes me nervous about edge cases.
> If you could use a single MPU to commit the final output, you get a nice O(1)
> atomic operation
Right now we write the flush or compaction output to a temporary directory,
then move them into place. This is, presumably, so if we crash in the midst of
the writing we don't then get an implicit close() that makes partial results
visible in the region data directory.
As an optimization we might try to write the file(s) into place with MPU and
instead of rename the commit operation is finalization of the MPU, as long as
S3 guarantees that if we crap out and don't finalize the results are never
visible, and as long as Ceph's radosgw supports all of the same behavior, which
is a big if...
Thanks for the input!
I would not expect you to volunteer code, no worries! (That would be
obnoxious... (smile))
> Store commit transaction for filesystems that do not support an atomic rename
> -----------------------------------------------------------------------------
>
> Key: HBASE-20431
> URL: https://issues.apache.org/jira/browse/HBASE-20431
> Project: HBase
> Issue Type: Sub-task
> Reporter: Andrew Purtell
> Priority: Major
>
> HBase expects the Hadoop filesystem implementation to support an atomic
> rename() operation. HDFS does. The S3 backed filesystems do not. The
> fundamental issue is the non-atomic and eventually consistent nature of the
> S3 service. A S3 bucket is not a filesystem. S3 is not always immediately
> read-your-writes. Object metadata can be temporarily inconsistent just after
> new objects are stored. There can be a settling period to ride over.
> Renaming/moving objects from one path to another are copy operations with
> O(file) complexity and O(data) time followed by a series of deletes with
> O(file) complexity. Failures at any point prior to completion will leave the
> operation in an inconsistent state. The missing atomic rename semantic opens
> opportunities for corruption and data loss, which may or may not be
> repairable with HBCK.
> Handling this at the HBase level could be done with a new multi-step
> filesystem transaction framework. Call it StoreCommitTransaction.
> SplitTransaction and MergeTransaction are well established cases where even
> on HDFS we have non-atomic filesystem changes and are our implementation
> template for the new work. In this new StoreCommitTransaction we'd be moving
> flush and compaction temporaries out of the temporary directory into the
> region store directory. On HDFS the implementation would be easy. We can rely
> on the filesystem's atomic rename semantics. On S3 it would be work: First we
> would build the list of objects to move, then copy each object into the
> destination, and then finally delete all objects at the original path. We
> must handle transient errors with retry strategies appropriate for the action
> at hand. We must handle serious or permanent errors where the RS doesn't need
> to be aborted with a rollback that cleans it all up. Finally, we must handle
> permanent errors where the RS must be aborted with a rollback during region
> open/recovery. Note that after all objects have been copied and we are
> deleting obsolete source objects we must roll forward, not back. To support
> recovery after an abort we must utilize the WAL to track transaction
> progress. Put markers in for StoreCommitTransaction start and completion
> state, with details of the store file(s) involved, so it can be rolled back
> during region recovery at open. This will be significant work in HFile,
> HStore, flusher, compactor, and HRegion. Wherever we use HDFS's rename now we
> would substitute the running of this new multi-step filesystem transaction.
> We need to determine this for certain, but I believe on S3 the PUT or
> multipart upload of an object must complete before the object is visible, so
> we don't have to worry about the case where an object is visible before fully
> uploaded as part of normal operations. So an individual object copy will
> either happen entirely and the target will then become visible, or it won't
> and the target won't exist.
> S3 has an optimization, PUT COPY
> (https://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectCOPY.html), which
> the AmazonClient embedded in S3A utilizes for moves. When designing the
> StoreCommitTransaction be sure to allow for filesystem implementations that
> leverage a server side copy operation. Doing a get-then-put should be
> optional. (Not sure Hadoop has an interface that advertises this capability
> yet; we can add one if not.)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
