Adar Dembo has posted comments on this change.

Change subject: KUDU-495 (part 1): new sys_catalog writing API
......................................................................


Patch Set 1:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/2648/1/src/kudu/master/sys_catalog.cc
File src/kudu/master/sys_catalog.cc:

Line 508: if
> nit, should we CHECK_OK here, like below? what do we do if we return corrup
I retained the existing semantics, which was RETURN_NOT_OK for serialization 
and CHECK_OK for row.SetFoo().

Once I'm done with my work, all callers will be able to handle failures in 
SysCatalogTable::Write() gracefully. That said, corruption is an interesting 
case because it's likely to repeat itself time after time. I don't know what we 
should do; seems like we need to either pretend that corruption is like any 
other failure and fail the client call, or we could kill the master. Neither 
are ideal.


http://gerrit.cloudera.org:8080/#/c/2648/1/src/kudu/master/sys_catalog.h
File src/kudu/master/sys_catalog.h:

Line 97: WriteActions
> this name is a bit strange. how about SysCatalogActions or even just action
I'll change it to Actions.


-- 
To view, visit http://gerrit.cloudera.org:8080/2648
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I7f5e5ecbff33435c964ced2e421566c41c387e9e
Gerrit-PatchSet: 1
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Adar Dembo <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: David Ribeiro Alves <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-HasComments: Yes

Reply via email to