You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@distributedlog.apache.org by "ASF GitHub Bot (JIRA)" <ji...@apache.org> on 2016/12/13 02:47:58 UTC
[jira] [Commented] (DL-94) Make the zookeeper client used by
bookkeeper client retry on session expires
[ https://issues.apache.org/jira/browse/DL-94?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15743935#comment-15743935 ]
ASF GitHub Bot commented on DL-94:
----------------------------------
GitHub user sijie opened a pull request:
https://github.com/apache/incubator-distributedlog/pull/67
DL-94: Make the zookeeper client used by bookkeeper client retry on session expires
merge twitter's change on session expires
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/sijie/incubator-distributedlog merge/DL-94
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/incubator-distributedlog/pull/67.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 #67
----
commit 904b8986b46fc908cfd1a0ee05b35dce77d12f5c
Author: Jordan Bull <jb...@twitter.com>
Date: 2016-05-13T18:27:19Z
when publishing per partition stats, also publish per stream stats
RB_ID=820062
commit 00919605f1bfa3f3e0c8514dc63552df4eeb3cc2
Author: Jordan Bull <jb...@twitter.com>
Date: 2016-05-18T17:18:34Z
use zero padded partition strings in stats
RB_ID=833803
commit 7b46a9ac6bb5d520366069c244332347ef019e8e
Author: Sijie Guo <si...@twitter.com>
Date: 2016-05-23T23:49:19Z
Improve handling of lock conflicts in zk session lock
- lock reacquire could happen in foreground and background thread. so use a semaphore to make sure there is only on e outstanding acquire operation. and check if already hold lock before reacquire.
- fix handling zk sibling znode logic. as the znode is sequential znode, the name would be different each time. so only comparing the client id and session id of the znodes
RB_ID=833945
commit 517c77c164cb989ae9829cbd80bf2e492eb8e364
Author: Leigh Stewart <ls...@twitter.com>
Date: 2016-12-13T00:33:33Z
Introduce periodic keepalive control record in writer
* so the writer will periodically write 'keepalive' control record to make sure the stream is alive. so if the write proxy is disconnected from bookies, the control record will fail to write hence write proxy gets a chance to drop the ownership.
commit b571d3b4adcc140acca881979782474c27459d34
Author: Sijie Guo <si...@twitter.com>
Date: 2016-05-24T04:01:57Z
DL: remove watches when unregister children watches
RB_ID=833858
commit 98dc9ab2df8af7af26ca68dbe93be3d420cda417
Author: Sijie Guo <si...@twitter.com>
Date: 2016-05-27T00:06:31Z
Remove watcher only when it registered watcher
(reduce the annoying logging from zookeeper client about "Failed to find watcher!"
RB_ID=837073
commit d3a97bc0dde0c25516840725599ac46fa03601ab
Author: Dave Rusek <dr...@twitter.com>
Date: 2016-06-06T23:50:25Z
dl: add flag to enable thrift mux on DL Client
RB_ID=839555
commit 800b867b705a786094b9a591e0ebc64fa934f632
Author: Dave Rusek <da...@gmail.com>
Date: 2016-12-13T00:41:47Z
Don't use stack and codec together for configuring thriftmux
- Don't use stack and codec together for configuring thriftmux
- Remove codec if the thriftmux is disabled
commit 0a18f564ba7e4d03f4c6a0859e5478535a59befd
Author: Leigh Stewart <ls...@twitter.com>
Date: 2016-12-13T00:46:11Z
Be able to close the writer within a timeout period
RB_ID=841340
commit f19e7564ff4a1ec1b5d6f2683db190d739df99bb
Author: Leigh Stewart <ls...@twitter.com>
Date: 2016-12-13T00:49:26Z
Simplify the state transition on stream
* the stream is created on INITIALIZING
* when the stream is started, it would start transition from INITIALIZING to INITIALIZED
* it would serve stream operations when the stream is INITIALIZED
* it would be turned to ERROR when encountered exceptions.
* the stream would be closed when service operation timeout or encountered any exceptions. it would first be removed from acquired mapping
* the stream would be removed from cached mapping depends on probation time.
RB_ID=848047
commit f008f751ec0a020a670903b01c709007ada87381
Author: Sijie Guo <si...@twitter.com>
Date: 2016-07-08T22:22:55Z
Allow configuring flush interval in micros for distributedlog multi streams writer
RB_ID=846974
commit 72a786e78540fb5a3f8e27f9d71d3616d06d1548
Author: Sijie Guo <si...@twitter.com>
Date: 2016-07-11T17:10:55Z
LogSegmentMetadataStore should only notify when the list of log segments is updated
Currently it notifies the listeners not only when there is a change but also when session expires. it would break the readahead loop and cause readers have to wait until it is able to connect to zookeeper again.
With this change, it would only notify when the list of log segments is updated. If it disconnects to zookeeper, the listener won't be notified and it would keep reading from the log segments it knows.
RB_ID=842998
commit f18fe172fb4559bcabb7173d1cd0b9d27fc23c93
Author: Sijie Guo <si...@twitter.com>
Date: 2016-07-11T17:21:44Z
Make the zookeeper client used by bookkeeper client retry on session expires
* the zookeeper client used by bookkeeper client is purely for metadata accesses, so we should retry on session expires.
* remove the unnessary zookeeper session handling in bk log handler. as we don't necessary to fail bookkeeper client or log handler when session expires as it would be handled and retried by the zookeeper client.
* Make the retry infinitely if the retry settings for bkc zookeeper client is set to 0 or negative.
RB_ID=843057
----
> Make the zookeeper client used by bookkeeper client retry on session expires
> ----------------------------------------------------------------------------
>
> Key: DL-94
> URL: https://issues.apache.org/jira/browse/DL-94
> Project: DistributedLog
> Issue Type: Improvement
> Components: distributedlog-core
> Reporter: Sijie Guo
> Assignee: Sijie Guo
> Fix For: 0.4.0
>
>
> * the zookeeper client used by bookkeeper client is purely for metadata accesses, so we should retry on session expires.
> * remove the unnessary zookeeper session handling in bk log handler. as we don't necessary to fail bookkeeper client or log
> handler when session expires as it would be handled and retried by the zookeeper client.
> * Make the retry infinitely if the retry settings for bkc zookeeper client is set to 0 or negative.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)