Adar Dembo has posted comments on this change.

Change subject: KUDU-1785. Fix potential crash in TabletCopySourceSession
......................................................................


Patch Set 2:

(6 comments)

http://gerrit.cloudera.org:8080/#/c/5363/2//COMMIT_MSG
Commit Message:

PS2, Line 14: Added a stress test in tablet_copy_client_session-itest to 
trigger the
            : issue.
How often was the issue triggered by this test without the fix?


http://gerrit.cloudera.org:8080/#/c/5363/2/src/kudu/integration-tests/tablet_copy_client_session-itest.cc
File src/kudu/integration-tests/tablet_copy_client_session-itest.cc:

PS2, Line 52:   // We don't want the master to interfere when we forcibly make 
copies of
            :   // tablets onto servers it doesn't know about. Only 2 servers.
Does this comment explain why master_tombstone_evicted_tablet_replicas=false?

If so, could you move it to be above that line too? If not, could you add a 
comment justifying the change from the non-default value?


Line 65:   workload.set_num_write_threads(8);
Why override the default value for this?


PS2, Line 70:   while (workload.rows_inserted() < 10000) {
            :     SleepFor(MonoDelta::FromMilliseconds(10));
            :   }
Why do we need to do this?


PS2, Line 88: StartSingleNode
Not sure this is an ideal name for the function; it's actually starting an 
entire cluster (master + 2 tservers). The magic is that all of the 
single-replica tablets are on the first tserver; maybe you can rename the 
function to make that explicit?


PS2, Line 109: std::numeric_limits<int32_t>::max()
Isn't this argument an int64_t though?

If it's optional, would be nice to use a boost::optional to convey that.


-- 
To view, visit http://gerrit.cloudera.org:8080/5363
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I6f3ec35885dbf1a81c23ac10b1c9556dfddbd4b7
Gerrit-PatchSet: 2
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Mike Percy <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Dinesh Bhat <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-HasComments: Yes

Reply via email to