Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/15828 )

Change subject: [master-test] added GetTableSchema() microbenchmarks
......................................................................


Patch Set 1:

(7 comments)

http://gerrit.cloudera.org:8080/#/c/15828/1//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/15828/1//COMMIT_MSG@13
PS1, Line 13: debug
> yea I think quoting release build numbers is probably more interesting
Added stats for RELEASE build as well.  Interestingly enough, in the case of 
RELEASE build the difference is huge for function call counters.


http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc
File src/kudu/master/master-test.cc:

http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc@1084
PS1, Line 1084: FLAGS_rpc_service_queue_length
> Probably, this should be min(FLAGS_rpc_service_queue_length, the number of
From the other side, it's desirable to test with as many threads as possible.  
The only requirement is to avoid RPC queue overflows.


http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc@1110
PS1, Line 1110: if (resp.has_error()) {
              :           LOG(WARNING) << SecureDebugString(resp);
              :           continue;
              :         }
> Do you think it's worth failing this test on error? That way we can more ea
Indeed, that's a better approach.  Done.


http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc@1137
PS1, Line 1137: 200
> Probably, this should depend on the number of CPU cores.
>From the other side, it's desirable to test with many threads each 
>representing a client calling GetTableSchema().


http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc@1139
PS1, Line 1139:   const bool supports_authz = GetParam();
              :   FLAGS_master_support_authz_tokens = supports_authz;
              :
              :   for (auto idx = 0; idx < kNumTables; ++idx) {
              :     auto table_name = Substitute("testtb_$0", idx);
              :     EXPECT_OK(CreateTable(table_name, kTableSchema));
              :   }
> nit: consider putting these in some common SetUp()?
Moved setting supports_authz_ into the constructor.

In different scenarios kNumTables is different, so SetUp() would not fit for 
the piece of the creation of tables, IMHO.


http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc@1180
PS1, Line 1180:   auto stopper_cleanup = MakeScopedCleanup([&]() {
              :     for (auto& t : caller_threads) {
              :       t.join();
              :     }
              :   });
> nit: just FYI, you can use this too if you'd like, especially if you don't
Thanks!  I always forget about those useful macros.


http://gerrit.cloudera.org:8080/#/c/15828/1/src/kudu/master/master-test.cc@1191
PS1, Line 1191: calls
> nit: maybe "function calls"? an RPC is also a "call", so it wasn't quite cl
Done



--
To view, visit http://gerrit.cloudera.org:8080/15828
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ic1f3d4fe1df426d50bd465c8ca7bbfb6924bffe9
Gerrit-Change-Number: 15828
Gerrit-PatchSet: 1
Gerrit-Owner: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Andrew Wong <[email protected]>
Gerrit-Reviewer: Grant Henke <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-Comment-Date: Tue, 28 Apr 2020 08:21:05 +0000
Gerrit-HasComments: Yes

Reply via email to