-
Notifications
You must be signed in to change notification settings - Fork 28.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[SPARK-4344][DOCS] adding documentation on spark.yarn.user.classpath.first #3209
Conversation
Can one of the admins verify this patch? |
@@ -224,6 +224,7 @@ Apart from these, the following properties are also available, and may be useful | |||
(Experimental) Whether to give user-added jars precedence over Spark's own jars when | |||
loading classes in Executors. This feature can be used to mitigate conflicts between | |||
Spark's dependencies and user dependencies. It is currently an experimental feature. | |||
(On Yarn use spark.yarn.user.classpath.first) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actually, the two config options do not work the same way. I'm working on implementing userClassPathFirst
for yarn mode and, in the process, will deprecate spark.yarn.user.classpath.first
, since the semantics of the former are less dangerous.
(If curious check out SPARK-2996, although the bug doesn't have a whole lot of information.)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Great - I think that will be clearer. Is it worth having a comment in 1.1 at least though that (correctly) states how and when to use which one?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Personally, I think spark.yarn.user.classpath.first
is a little too dangerous for us to publicly document - and we've lived so far without documenting it, and asking people to try it on a as-needed basis. But let's see if others (@tgravescs @andrewor14) have a different opinion.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
it was left undocumented on purpose (expert config) used as last resort type thing.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
OK - perhaps than just a note that the other configuration won't work on Yarn?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah I think it's sufficient to add a one line doc that the other config doesn't work on Yarn (have you verified this?)
821862b
to
51cb9b2
Compare
Updated it to the one line change on userClassPathFirst, but just as happy to close it if #3233 goes in |
I'm ok with the message, although if you only add it on branch-1.2 it would save me the work of having to delete it as part of my other PR. :-) |
I don't know if it's possible to move a PR to a different branch (or whether you need to create a new one). In any case, it's not a big deal if this goes into master. |
test this please |
I pulled this into both master and branch-1.2 |
…first The documentation for the two parameters is the same with a pointer from the standalone parameter to the yarn parameter Author: arahuja <aahuja11@gmail.com> Closes #3209 from arahuja/yarn-classpath-first-param and squashes the following commits: 51cb9b2 [arahuja] [SPARK-4344][DOCS] adding documentation for YARN on userClassPathFirst (cherry picked from commit d240760) Signed-off-by: Thomas Graves <tgraves@apache.org>
The documentation for the two parameters is the same with a pointer from the standalone parameter to the yarn parameter