Adar Dembo has posted comments on this change.

Change subject: KUDU-1793: only update lbm container bookkeeping on success
......................................................................


Patch Set 3:

(9 comments)

http://gerrit.cloudera.org:8080/#/c/5399/3/src/kudu/fs/block_manager-test.cc
File src/kudu/fs/block_manager-test.cc:

PS3, Line 1160: kNumAppendTries
> nit: sounds like multiple attempts/retries at the same operation. Maybe kNu
Done


PS3, Line 1168: kTestdata
> nit: kTestData
Done


PS3, Line 1179: sizeof
> nit: arraysize is better than sizeof
Doesn't seem to work:

/home/adar/Source/kudu/src/kudu/fs/block_manager-test.cc:1191:92: error: no 
matching function for call to ‘ArraySizeHelper(uint8_t [(<anonymous> + 1)])’
In file included from /home/adar/Source/kudu/src/kudu/fs/block_id.h:26:0,
                 from 
/home/adar/Source/kudu/src/kudu/fs/file_block_manager.h:26,
                 from 
/home/adar/Source/kudu/src/kudu/fs/block_manager-test.cc:24:
/home/adar/Source/kudu/src/kudu/gutil/macros.h:142:8: note: candidate: 
template<class T, long unsigned int N> char (& ArraySizeHelper(T (&)[N]))[N]
 char (&ArraySizeHelper(T (&array)[N]))[N];
        ^
/home/adar/Source/kudu/src/kudu/gutil/macros.h:142:8: note:   template argument 
deduction/substitution failed:
/home/adar/Source/kudu/src/kudu/fs/block_manager-test.cc:1191:92: note:   
variable-sized array type ‘long int’ is not a valid template argument
In file included from /home/adar/Source/kudu/src/kudu/fs/block_id.h:26:0,
                 from 
/home/adar/Source/kudu/src/kudu/fs/file_block_manager.h:26,
                 from 
/home/adar/Source/kudu/src/kudu/fs/block_manager-test.cc:24:
/home/adar/Source/kudu/src/kudu/gutil/macros.h:149:8: note: candidate: 
template<class T, long unsigned int N> char (& ArraySizeHelper(const T 
(&)[N]))[N]
 char (&ArraySizeHelper(const T (&array)[N]))[N];
        ^
/home/adar/Source/kudu/src/kudu/gutil/macros.h:149:8: note:   template argument 
deduction/substitution failed:
/home/adar/Source/kudu/src/kudu/fs/block_manager-test.cc:1191:92: note:   
variable-sized array type ‘long int’ is not a valid template argument


PS3, Line 1179: 0 
> erm?
Ugh, that should have been 'i', not 0.


http://gerrit.cloudera.org:8080/#/c/5399/3/src/kudu/fs/log_block_manager.cc
File src/kudu/fs/log_block_manager.cc:

PS3, Line 206:  next_append_bytes
> I find this parameter name a little confusing. maybe next_append_size or ne
Will use next_append_length.


Line 672:     
RETURN_NOT_OK(data_dir_->RefreshIsFull(DataDir::RefreshMode::ALWAYS));
> not a new issue, but question here: if preallocation is not enabled, does t
Yes, if you've disabled preallocation, you're getting a statvfs() with every 
write.

As for whether or not it's a big deal, I don't know. I'd expect that 
information to be cached, so it'd be the overhead of an extra syscall at worst. 
I haven't fretted about it because disabling preallocation just seems like a 
bad idea in general.


http://gerrit.cloudera.org:8080/#/c/5399/3/src/kudu/util/env_posix.cc
File src/kudu/util/env_posix.cc:

PS3, Line 110: file_allocation_injected_failure
> the other injection flags follow a pattern more like 'env_inject_error_on_f
Will rename to env_inject_io_error_on_write_or_preallocate.


PS3, Line 371: RuntimeError
> I think an IOError would be more consistent with the status type that is re
Done


Line 576:     MAYBE_RETURN_FAILURE(FLAGS_file_allocation_injected_failure,
> the name here is not quite matching up. Perhaps the flag should be clearer 
I think the new name (see earlier comment) makes it clear.


-- 
To view, visit http://gerrit.cloudera.org:8080/5399
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I49bc98c9f8b7dce0333f88cec85757fe122acfa4
Gerrit-PatchSet: 3
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Adar Dembo <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Dan Burkert <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-HasComments: Yes

Reply via email to