impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tim Armstrong (Code Review)" <>
Subject [Impala-ASF-CR] IMPALA-4114: Port BufferedBlockMgr tests to buffer pool
Date Fri, 07 Apr 2017 15:49:11 GMT
Tim Armstrong has posted comments on this change.

Change subject: IMPALA-4114: Port BufferedBlockMgr tests to buffer pool

Patch Set 6:

File be/src/runtime/bufferpool/

PS6, Line 1434: flushing heuristic does not eagerly start writes when there are no
              :   // disks.
> which path in WriteDirtyPagesAsync() is this?
const int64_t target_writes = FLAGS_concurrent_scratch_ios_per_device
      * file_group_->tmp_file_mgr()->NumActiveTmpDevices();

works out to be 0, because there are no active temporary devices.

Line 1479:   IMPALA_ASSERT_DEBUG_DEATH(pool->Unpin(&client, &pages[0]), "");
> just to confirm, this is the DCHECK_EQ(0, dirty_unpinned_bytes_); in WriteD
It's DCHECK(spilling_enabled()). This is a little subtle - this test is setting up the client
via a QueryState(), so the logic in QueryState about scratch_limit is enabled. Updated comment
to mention this.
File be/src/runtime/

PS5, Line 274: tus err_status(TErrorCode::SCRATCH_ALLOCATION_FAILED);
             :     for (Status& err : scratch_errors_) err_status.Merg
> isn't that still the case? will we want to update this message still when 4
Absolutely. There was a pre-existing copy of this comment in common/thrift/
next to the message. I think what happened was that I had two duplicate errors in this file,
so intended to consolidate them in, but only moved one of them over.
File be/src/runtime/

Line 358:   LOG(INFO) << "Allocated in " << tmp_file->path();
> did you mean to keep this?

To view, visit
To unsubscribe, visit

Gerrit-MessageType: comment
Gerrit-Change-Id: Ifb0221e8bea6f3b23b62d5094634d97562295ea3
Gerrit-PatchSet: 6
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Tim Armstrong <>
Gerrit-Reviewer: Dan Hecht <>
Gerrit-Reviewer: Tim Armstrong <>
Gerrit-HasComments: Yes

View raw message