Andrew Wong has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/13426 )

Change subject: KUDU-2797: the master aggregates tablet statistics
......................................................................


Patch Set 11:

(1 comment)

Haven't looked in depth, but I left feedback on the new approach.

http://gerrit.cloudera.org:8080/#/c/13426/11/src/kudu/tablet/tablet_replica.cc
File src/kudu/tablet/tablet_replica.cc:

http://gerrit.cloudera.org:8080/#/c/13426/11/src/kudu/tablet/tablet_replica.cc@835
PS11, Line 835:     
pb.set_rows_inserted(tablet->metrics()->rows_inserted->value());
              :     
pb.set_rows_upserted(tablet->metrics()->rows_upserted->value());
              :     
pb.set_rows_updated(tablet->metrics()->rows_updated->value());
              :     
pb.set_rows_deleted(tablet->metrics()->rows_deleted->value());
              :     pb.set_insertions_failed_dup_key(
              :         tablet->metrics()->insertions_failed_dup_key->value());
              :     
pb.set_upserts_as_updates(tablet->metrics()->upserts_as_updates->value());
              :     
pb.set_scanner_rows_returned(tablet->metrics()->scanner_rows_returned->value());
              :     
pb.set_scanner_cells_returned(tablet->metrics()->scanner_cells_returned->value());
              :     
pb.set_scanner_bytes_returned(tablet->metrics()->scanner_bytes_returned->value());
              :     
pb.set_scanner_rows_scanned(tablet->metrics()->scanner_rows_scanned->value());
              :     pb.set_scanner_cells_scanned_from_disk(
              :         
tablet->metrics()->scanner_cells_scanned_from_disk->value());
              :     pb.set_scanner_bytes_scanned_from_disk(
              :         
tablet->metrics()->scanner_bytes_scanned_from_disk->value());
              :     
pb.set_scans_started(tablet->metrics()->scans_started->value());
              :     
pb.set_bloom_lookups(tablet->metrics()->bloom_lookups->value());
              :     
pb.set_key_file_lookups(tablet->metrics()->key_file_lookups->value());
              :     
pb.set_delta_file_lookups(tablet->metrics()->delta_file_lookups->value());
              :     pb.set_mrs_lookups(tablet->metrics()->mrs_lookups->value());
              :     
pb.set_bytes_flushed(tablet->metrics()->bytes_flushed->value());
              :     pb.set_undo_delta_block_gc_bytes_deleted(
              :         
tablet->metrics()->undo_delta_block_gc_bytes_deleted->value());
              :     pb.set_mem_rowset_size(tablet->MemRowSetSize());
              :     pb.set_on_disk_data_size(tablet->OnDiskDataSize());
              :     pb.set_num_rowsets(tablet->num_rowsets());
              :     pb.set_on_disk_size(OnDiskSize());
              :     pb.set_live_row_count(live_row_count);
I'd advocate for _not_ sending back every single one of these metrics, and 
limiting the scope of TabletStatsPB to just live row count and on-disk size. 
AFAICT the goals of this patch are to:
- make it convenient to get the size of a table, which is a pretty common task
- lay the groundwork for more complex statistics

I worry that sending _all_ tablet metrics like this will:
- not be useful very frequently (beyond just data size), e.g. it doesn't seem 
particularly useful to know about the number of rowsets across a table, vs in a 
single tablet replica, so I'm not sure if there's much value added for 
aggregating them on the master
- be somewhat expensive if sent for thousands of tablets. Plus, since we're 
sending so many metrics back, this will likely make tablets dirty way more 
frequently (e.g. a tablet will be dirty if it gets scanned, if it compacts at 
all, if it GCs WALs, etc.)



--
To view, visit http://gerrit.cloudera.org:8080/13426
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I74406ab7cca7c22fda455c328b8ee9989a6b2d99
Gerrit-Change-Number: 13426
Gerrit-PatchSet: 11
Gerrit-Owner: helifu <hzhel...@corp.netease.com>
Gerrit-Reviewer: Adar Dembo <a...@cloudera.com>
Gerrit-Reviewer: Andrew Wong <aw...@cloudera.com>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Mike Percy <mpe...@apache.org>
Gerrit-Reviewer: helifu <hzhel...@corp.netease.com>
Gerrit-Comment-Date: Wed, 19 Jun 2019 16:58:26 +0000
Gerrit-HasComments: Yes

Reply via email to