[
https://issues.apache.org/jira/browse/HDFS-5138?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13870417#comment-13870417
]
Aaron T. Myers commented on HDFS-5138:
--------------------------------------
Thanks for the comments, Konst.
bq. This is less intuitive than the current state of the art. Because after NN
rollback you need to start NameNode as -regular, while DataNodes with -rollback
startup option.
It's different, but not obvious to me that it's necessary less intuitive. I've
personally always found it a bit strange that to roll back you need to start
the NN _once_ with the '-rollback' option, which will result in it doing some
things at startup, and then starting up as normal. This might seem to imply
that the NN is running in some sort of "rollback mode", when in fact the act of
rolling back has already completed, and thereafter you should always start the
NN without the '-rollback' option.
bq. Also just mentioning there could be some collisions with the rolling
upgrade design, which I just finished reading. I think HDFS-5535 assumes
current (pre-your-patch) behaviours of -rollback and -finalize. For -finalize
the problem could be that you remove it as a start up option. May be Suresh can
elaborate better on this.
Needing to roll back should (hopefully!) be such a rare occurrence that it
doesn't seem unreasonable to me to not do that in a rolling way. Removal of the
'-finalize' startup option, I would think, should make the whole thing easier,
and doesn't seem to me to have any benefits vs. just using the finalizeUpgrade
RPC.
> Support HDFS upgrade in HA
> --------------------------
>
> Key: HDFS-5138
> URL: https://issues.apache.org/jira/browse/HDFS-5138
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.1.1-beta
> Reporter: Kihwal Lee
> Assignee: Aaron T. Myers
> Priority: Blocker
> Attachments: HDFS-5138.patch, HDFS-5138.patch, HDFS-5138.patch,
> HDFS-5138.patch, HDFS-5138.patch, HDFS-5138.patch, HDFS-5138.patch
>
>
> With HA enabled, NN wo't start with "-upgrade". Since there has been a layout
> version change between 2.0.x and 2.1.x, starting NN in upgrade mode was
> necessary when deploying 2.1.x to an existing 2.0.x cluster. But the only way
> to get around this was to disable HA and upgrade.
> The NN and the cluster cannot be flipped back to HA until the upgrade is
> finalized. If HA is disabled only on NN for layout upgrade and HA is turned
> back on without involving DNs, things will work, but finaliizeUpgrade won't
> work (the NN is in HA and it cannot be in upgrade mode) and DN's upgrade
> snapshots won't get removed.
> We will need a different ways of doing layout upgrade and upgrade snapshot.
> I am marking this as a 2.1.1-beta blocker based on feedback from others. If
> there is a reasonable workaround that does not increase maintenance window
> greatly, we can lower its priority from blocker to critical.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
