[
https://issues.apache.org/jira/browse/BOOKKEEPER-1086?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16050856#comment-16050856
]
ASF GitHub Bot commented on BOOKKEEPER-1086:
--------------------------------------------
GitHub user athanatos opened a pull request:
https://github.com/apache/bookkeeper/pull/193
BOOKKEEPER-1086: ZkUnderreplicationManager cache watcher
Previously, getLedgerToReplicate left watches each time it traversed the
tree until it found a suitable replication target. Since we don't have
a way of canceling watches, these watches tended to get abandoned,
particularly on interior nodes, which aren't changed much. Thus,
over time, some nodes would build up a very large number of watch.
Instead, introduce a caching mechanism to remember outstanding watches
and avoid ever creating two watches on the same node.
Author: Samuel Just <sj...@salesforce.com>
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/athanatos/bookkeeper
forupstream/BOOKKEEPER-1098
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/bookkeeper/pull/193.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #193
----
commit cb3c7a2f4e256de00268740c37fc2a2ecd9b4e31
Author: Samuel Just <sj...@salesforce.com>
Date: 2017-03-29T19:06:56Z
BOOKKEEPER-1086: ZkUnderreplicationManager cache watcher
Previously, getLedgerToReplicate left watches each time it traversed the
tree until it found a suitable replication target. Since we don't have
a way of canceling watches, these watches tended to get abandoned,
particularly on interior nodes, which aren't changed much. Thus,
over time, some nodes would build up a very large number of watch.
Instead, introduce a caching mechanism to remember outstanding watches
and avoid ever creating two watches on the same node.
Author: Samuel Just <sj...@salesforce.com>
----
> Ledger Recovery - Refactor PendingReadOp
> ----------------------------------------
>
> Key: BOOKKEEPER-1086
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-1086
> Project: Bookkeeper
> Issue Type: Sub-task
> Components: bookkeeper-client
> Reporter: Sijie Guo
> Assignee: Sijie Guo
> Fix For: 4.5.0
>
>
> bookkeeper recovery improvement (part-1): refactor PendingReadOp
> this change is the first part of improving bookkeeper recovery. it is
> basically a refactor change, which:
> - abstract an interface for LedgerEntryRequest in PendingReadOp
> - rename current implementation to SequenceReadRequest, which read the
> entry in the sequence of quorum.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
