hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ray Chiang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-6142) Support rolling upgrade between 2.x and 3.x
Date Thu, 14 Sep 2017 17:32:00 GMT

    [ https://issues.apache.org/jira/browse/YARN-6142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16166655#comment-16166655
] 

Ray Chiang commented on YARN-6142:
----------------------------------

h2. Proto Changes

h3. ResourceTracker.proto
* Added unRegisterNodeManager.  Change due to YARN-41.  Confirmed in 2.8.0 and later.
** Potential issue with upgraded NM but not RM.  => RM should be upgraded first.

h3. RPCHeader.proto
* Not related to YARN.  Looks like support for HTrace.

h3. applicationclient_protocol.proto
* Added failApplicationAttempt
** YARN-261.  Confirmed in 2.8.0 and later.
* Added getNewReservation
** YARN-4957.  Confirmed in 2.8.0 and later.
* Added listReservations
** YARN-4340.  Confirmed in 2.8.0 and later.
* Added updateApplicationPriority
** YARN-4014.  Confirmed in 2.8.0 and later.
* Added signalToContainer
** MAPREDUCE-5044.  Confirmed in 2.8.0 and later.
* Added updateApplicationTimeouts
** YARN-5611.  Confirmed in 2.9.0 and later.

h3. collectornodemanager_protocol.proto
* Requires Timeline Service to be running first?

h3. containermanagement_protocol.proto
* Added increaseContainersResource
** YARN-1449.  2.8.0 and later
* Added updateContainer
** YARN-5977.  2.9.0 and later
* signalToContainer
** MAPREDUCE-5044.  Confirmed 2.8.0 and later.
* Added localizeContainer 
** YARN-5557.  2.9.0 and later
* Added reInitializeContainer/restartContainer/rollbackContainer/commitContainer
** YARN-5609.  2.9.0 and later

h3. distributed_scheduling_am_protocol.proto
* All for distributed scheduling support in the AM

h3. mr_protos.proto
* Two optional fields added.

h3. resourcemanager_adminstration_protocol.proto
* All methods for RMAdmin CLI

h3. yarn_security_token.proto
* Four optional fields added
* Added nodeLabelExpression
** YARN-3354.  Confirmed 2.8.0 and later.
* Added containerType
** YARN-3116.  Confirmed 2.8.0 and later.
* Added executionType
** YARN-2882.  Confirmed 2.9.0 and later.
* Added version
** YARN-5221.  Confirmed 2.8.0 and later.

h3. yarn_server_common_protos.proto
* Four optional fields added

h3. yarn_server_nodemanager_recovery.proto
* All optional fields and self-contained message changes

h3. yarn_server_resourcemanager_recovery.proto
* All optional fields and self-contained message changes



h2. New proto files (i.e. ignorable)
* yarn_server_federation_protos.proto


h2. Still to complete
* Security.proto
* yarn_protos.proto
* yarn_server_common_service_protos.proto
* yarn_server_resourcemanager_service_protos.proto
* yarn_service_protos.proto


> Support rolling upgrade between 2.x and 3.x
> -------------------------------------------
>
>                 Key: YARN-6142
>                 URL: https://issues.apache.org/jira/browse/YARN-6142
>             Project: Hadoop YARN
>          Issue Type: Task
>          Components: rolling upgrade
>    Affects Versions: 3.0.0-alpha2
>            Reporter: Andrew Wang
>            Assignee: Ray Chiang
>            Priority: Blocker
>
> Counterpart JIRA to HDFS-11096. We need to:
> * examine YARN and MR's  JACC report for binary and source incompatibilities
> * run the [PB differ|https://issues.apache.org/jira/browse/HDFS-11096?focusedCommentId=15816405&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15816405]
that Sean wrote for HDFS-11096 for the YARN PBs.
> * sanity test some rolling upgrades between 2.x and 3.x. Ideally these are automated
and something we can run upstream.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org


Mime
View raw message