-
Notifications
You must be signed in to change notification settings - Fork 8.7k
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
HADOOP-19161. S3A: option "fs.s3a.performance.flags" to take list of performance flags #6789
base: trunk
Are you sure you want to change the base?
HADOOP-19161. S3A: option "fs.s3a.performance.flags" to take list of performance flags #6789
Conversation
note the commented out bit where we considered adding options like "hive" or "spark". @HarshitGupta11 and I discussed this; for now lets go with a list of options and "*" |
/* case "hive": | ||
case "impala": | ||
case "spark": | ||
case "distcp": |
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.
Should we not let downstreamers decide what flags they want (after extensive testing)? And across different releases, they might need different flags to be turned on (in case of any regression)?
We can just recommend the flags (as already commented out here) but not set the flags for them. Thoughts?
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.
harshit and I were discussing this. i think it's best to have that option list, as app settings could be too brittle to changes
public boolean isDelete() { | ||
return delete; | ||
} |
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.
This one also we want to tackle as separate task (after HADOOP-19072), correct?
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.
yes. harshit did an experiment where he turned off all attempts at creating parent dirs after delete. fairly brittle, i think
I have a better design for this. changign this to draft. Proposed: we have a this makes it trivial to reuse/process. the implementation would be outside the actual Configuration class to make it easy for AbfsConfiguration to use too |
feb384f
to
28cd50a
Compare
|
28cd50a
to
f3571ac
Compare
💔 -1 overall
This message was automatically generated. |
…performance flags * A new FlagSet class in hadoop common enables this * and Configuration.getEnumSet() supports getting a set of enum values. * served up via StoreContext. Not sure about the merits of that I think it is needed so it gets down to all AbstractStoreOperation instances, * create performance is wired up. * tests which configure fs.s3a.create.performance clear fs.s3a.performance.flags in test setup. Change-Id: I52e48d19c624e7c18f22b3130943ffe72fac501f
a564aa4
to
82974c4
Compare
💔 -1 overall
This message was automatically generated. |
HADOOP-19161
Initial design
For testing we need to make sure ths is unset from all cost tests.
relates to #6543; the logic to set up that operation is here...that PR would
just be the implementation.
Same for a delete optimisation where we'd skip parent dir probe.
rename could do the same for its source dir too.
For code changes:
LICENSE
,LICENSE-binary
,NOTICE-binary
files?