Marcel Kornacker has posted comments on this change.

Change subject: IMPALA-2979: Fix scheduling on remote hosts
......................................................................


Patch Set 25:

(20 comments)

the header file looks good structurally

http://gerrit.cloudera.org:8080/#/c/2200/24/be/src/scheduling/simple-scheduler.h
File be/src/scheduling/simple-scheduler.h:

Line 380:       const TResourceBrokerReservationResponse& reservation, 
Coordinator* coord);
> Done. Being a class member method it will also affect member variables of t
the comments needs to describe the externally visible behavior, which includes 
side effects. as part of that, you should point out which class variables get 
updated (unless the answer is 'most of them').


http://gerrit.cloudera.org:8080/#/c/2200/25/be/src/scheduling/simple-scheduler.h
File be/src/scheduling/simple-scheduler.h:

Line 91:   /// List of backends.
no need to repeat/paraphrase type name


Line 112:     /// Construct maps from list of backends.
config


Line 137:   /// Internal structure to track scan range assignments for each 
backend and prioritize
"for a backend". remove comment about prioritization, this structure doesn't do 
the prioritization.


Line 141:     /// The number of bytes assigned to a backend host. Part of the 
key of
backend or backend host?


Line 142:     /// AddressableAssignmentHeap. This gets initialized when 
elements are inserted into
don't comment on other data structures here


Line 164:   /// will make the top() element of the heap be the backend with the 
least number of
"lowest number"


Line 174:   /// update their key. For each plan node we create a new heap, so 
they are not shared
the last sentence should go into the structure that contains an object of this 
class


Line 199:   /// member in AssignmentCtx.
move into AssignmentCtx? is it used outside of that?


Line 222:     /// data_locations:     List of candidate replica IP addresses, 
from which the block
redundant w/ earlier part of comment


Line 224:     /// break_ties_by_rank: Whether to consider the random rank 
during backend selection.
same here


Line 229:     /// those will be preferred. Otherwise one is selected from the 
assignment heap.
describe the externally observable behavior (what are the characteristics of 
the selected backend), not the implementation ("from assignment heap")


Line 233:     /// pointer. This assumes that a returned backend will also be 
assigned to. The caller
"increment the internal pointer" isn't meaningful


Line 242:         const TScanRangeLocations& scan_range_locations,
role of this param?


Line 252:     /// Used to lookup hostnames to IP addresses and IP addresses to 
backend.
use verb, not noun


Line 291:   /// The scheduler's backend maps. When receiving changes to the 
backend configuration
replace use of term 'backend maps' throughout


Line 292:   /// from the statestore we will make a copy of those maps, apply 
the updates to the copy
this is fairly central to how this class functions, and should be pointed out 
in the class comment


Line 424:   /// assigned work is picked, thus aiming to distribute the work as 
even as possible.
as evenly


Line 426:   /// Finally scan ranges are considered, which do not have an 
impalad backend running on
remove comma


Line 446:   ///   amount of assigned work), then the first one will be picked 
deterministically.
only true for disk replicas, no?


-- 
To view, visit http://gerrit.cloudera.org:8080/2200
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I044f83806fcde820fcb38047cf6b8e780d803858
Gerrit-PatchSet: 25
Gerrit-Project: Impala
Gerrit-Branch: cdh5-trunk
Gerrit-Owner: Lars Volker <[email protected]>
Gerrit-Reviewer: Lars Volker <[email protected]>
Gerrit-Reviewer: Marcel Kornacker <[email protected]>
Gerrit-Reviewer: Matthew Jacobs <[email protected]>
Gerrit-Reviewer: Mostafa Mokhtar <[email protected]>
Gerrit-Reviewer: Sailesh Mukil <[email protected]>
Gerrit-Reviewer: anujphadke <[email protected]>
Gerrit-HasComments: Yes

Reply via email to