Charles Connell created HDFS-18017:
--------------------------------------
Summary: Consolidate over-replicated erasure-coded stripes
Key: HDFS-18017
URL: https://issues.apache.org/jira/browse/HDFS-18017
Project: Hadoop HDFS
Issue Type: Sub-task
Components: erasure-coding
Affects Versions: 3.4.3
Reporter: Charles Connell
Assignee: Charles Connell
Relief copies that landed in domains without room stay as protective
duplicates: for example two domains at their budget share a copy of one
internal block while a third domain has room. Neither shared copy can be
removed alone, but copying that internal block into the roomy domain lets both
be removed, a net reduction of one.
Excess handling now chooses all removable copies of an internal block together,
since after such a copy the two old copies may only be removable as a pair. A
group left with protective duplicates that a single copy would reduce is queued
for consolidation. The redundancy monitor serves that queue only with capacity
left over from reconstruction, replanning under the lock before scheduling, so
a group is never less safe than before at any step.
AI disclosure: I used AI assist in researching this issue, writing the
forthcoming PR, and writing this issue description.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]