[
https://issues.apache.org/jira/browse/HBASE-12332?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14241478#comment-14241478
]
Jonathan Hsieh commented on HBASE-12332:
----------------------------------------
Do we have a test that shows that the mechanism you propose works when have a
mob hfile move from the mob dir to the archive dir.
Here's where I believe the we'll run into a problem
1) create a table with mobs,
2) snapshot it,
3) clone/restore it as a a different table
4) have a read workload on the snapshot
5) delete the original table
The problem is that if we are in the middle of reading a mob file and it is
moved to the archive, the read could fail and it is not clear if they read
would be retried against the new location from the same spot.
The hfile link would transparently fall over to alternate fie locations and
continue the read from the same place it had encountered the exception on the
new location. This avoids having to percolate an exception up as high.
>From a quick scan of the patch, I believe it covers the case where reads are
>not active or if a mob file move happens while reads are not in flight. I
>didn't see the case I described being covered.
> [mob] use filelink instad of retry when resolving an hfilelink.
> ---------------------------------------------------------------
>
> Key: HBASE-12332
> URL: https://issues.apache.org/jira/browse/HBASE-12332
> Project: HBase
> Issue Type: Sub-task
> Components: mob
> Affects Versions: hbase-11339
> Reporter: Jonathan Hsieh
> Fix For: hbase-11339
>
> Attachments: HBASE-12332-V1.diff
>
>
> in the snapshot code, hmobstore was modified to traverse an hfile link to a
> mob. Ideally this should use the transparent filelink code to read the data.
> Also there will likely be some issues with the mob file cache with these
> links.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
