Todd Lipcon has posted comments on this change.

Change subject: tools: add a tool to summarize data size on a tablet or tablets
......................................................................


Patch Set 2:

(14 comments)

http://gerrit.cloudera.org:8080/#/c/5727/2/src/kudu/tools/kudu-tool-test.cc
File src/kudu/tools/kudu-tool-test.cc:

Line 387:     const vector<string> kLocalReplicaModeRegexes = {
> Update this?
Done


Line 1119:             "KuduTableTestId | test-tablet | 0         | c10 (key)   
     | \\S+\n",
> Why are these column IDs c10, c11, c12 and not c1, c2, c3? AFAICT the schem
in debug builds the column ids start at 10. I'm surprised this didn't fail in a 
release build, though... my guess is that the '|' in these patterns is getting 
interpreted as 'OR', so the assertions are actually whack. I futzed with regex 
escaping for a while and eventually gave up and went with a new approach.


Line 1143:                 " KuduTableTestId | *           | *         | *      
          | \\S+\n"}) {
> Nit: fix indentation? Looks like some rows are written differently than oth
my editor doesn't like multiline strings


http://gerrit.cloudera.org:8080/#/c/5727/2/src/kudu/tools/tool_action_common.h
File src/kudu/tools/tool_action_common.h:

Line 65: extern const char* const kTabletIdGlobArg;
> You could define these just once in tool_action_local_replica.cc, since the
Done


http://gerrit.cloudera.org:8080/#/c/5727/1/src/kudu/tools/tool_action_local_replica.cc
File src/kudu/tools/tool_action_local_replica.cc:

PS1, Line 375:                      const vector<BlockId>& blocks,
             :                      StringPiece block_type,
> Would be nice to relegate this to "--verbose" or some such.
did check against VLOG(1)


Line 381:                           Substitute("could not open block $0", 
b.ToString()));
> As we discussed, it would also be nice to be able to sum up redo or undo or
done


Line 411: 
> This one too.
Done


http://gerrit.cloudera.org:8080/#/c/5727/2/src/kudu/tools/tool_action_local_replica.cc
File src/kudu/tools/tool_action_local_replica.cc:

Line 385:     *running_sum += size;
> Nit: let's only update the OUT param on success.
Done


Line 386:     if (VLOG_IS_ON(1)) {
> Do you want to piggy-back on the CLI's existing --verbose flag for this?
I don't see such a flag:

todd@va1022:~/kudu$ ./build/latest/bin/kudu --verbose
ERROR: unknown command line flag 'verbose'


Line 387:       cout << block_type << " block " << b.ToString() << ": "
> Nit: Substitute() so this is more readable?
Done


Line 447:     if (!MatchPattern(tablet_id, tablet_id_pattern)) continue;
> Should probably fully document the pattern matching semantics in the tool's
given that this is on tablet IDs I think really only '*' is particularly useful 
(so you can just type a prefix and use '*' to complete it). Tweaked the help a 
bit though.


Line 471:             (col_idx != Schema::kColumnNotFound) ?
> Is that possible? For a dropped column? Would be nice to get some test cove
yea for dropped. I think in this context it's a bit tricky to engineer the 
dropped column (no good utility code to do so).


Line 969:       .Description("Summarize the data size/space usage of the given 
local replica(s).")
> It would also be nice to slice the data vertically. For example, I want to 
It already does show the sum of redo/undo/bloom/pk/each-col across each table. 
Summing across all tables isn't done but how about leaving that for a future 
enhancement?


Line 982:       .AddAction(std::move(data_size))
> Nit: resort alphabetically.
Done


-- 
To view, visit http://gerrit.cloudera.org:8080/5727
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Ibf0237f9d01f4ec332d1df545d2c2f51f64ce012
Gerrit-PatchSet: 2
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Todd Lipcon <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Jean-Daniel Cryans <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <[email protected]>
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-Reviewer: Will Berkeley <[email protected]>
Gerrit-HasComments: Yes

Reply via email to