[feature](cloud) Add lazy commit mechanism for commit_txn#38243
Merged
gavinchou merged 5 commits intoapache:masterfrom Sep 1, 2024
Merged
[feature](cloud) Add lazy commit mechanism for commit_txn#38243gavinchou merged 5 commits intoapache:masterfrom
commit_txn#38243gavinchou merged 5 commits intoapache:masterfrom
Conversation
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
Contributor
Author
|
run buildall |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
0cac9c5 to
d47f33a
Compare
Contributor
Author
|
run buildall |
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
d47f33a to
a7e2b1b
Compare
Contributor
Author
|
run buildall |
1 similar comment
Contributor
Author
|
run buildall |
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
2 times, most recently
from
f70c9f4 to
d7c4b5c
Compare
Contributor
Author
|
run buildall |
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
d7c4b5c to
bf6aad6
Compare
Contributor
Author
|
run buildall |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
bf6aad6 to
9511fed
Compare
Contributor
Author
|
run buildall |
1 similar comment
Contributor
Author
|
run buildall |
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
w41ter
reviewed
Jul 26, 2024
| } | ||
|
|
||
| err = txn->commit(); | ||
| if (err != TxnErrorCode::TXN_OK) { |
Contributor
There was a problem hiding this comment.
There may be conflicts caused by concurrent commits. According to our design expectations, TXN_CONFLICT also needs to be handled. One simple solution is to consider the data already converted if TXN_CONFLICT occurs (this requires that each task uses the same split batch method). Another solution is to retry once if TXN_CONFLICT occurs.
It's essential to manage the competition between transactions properly here.
w41ter
reviewed
Jul 26, 2024
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
9511fed to
8ed4422
Compare
Contributor
Author
|
run buildall |
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
8ed4422 to
ae48134
Compare
Contributor
Author
|
run buildall |
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
ae48134 to
f23de13
Compare
Contributor
Author
|
run buildall |
swjtu-zhanglei
force-pushed
the
master-large-txn-new
branch
from
f23de13 to
c0a6c5a
Compare
TPC-DS: Total hot run time: 191755 ms |
ClickBench: Total hot run time: 31.14 s |
Contributor
Author
|
run buildall |
| std::unique_ptr<Transaction>& txn, MetaServiceCode& code, | ||
| std::string& msg); | ||
|
|
||
| void convert_tmp_rowsets( |
Contributor
There was a problem hiding this comment.
warning: function 'convert_tmp_rowsets' exceeds recommended size/complexity thresholds [readability-function-size]
void convert_tmp_rowsets(
^Additional context
cloud/src/meta-service/txn_lazy_committer.cpp:40: 154 lines including whitespace and comments (threshold 80)
void convert_tmp_rowsets(
^| } | ||
| } | ||
|
|
||
| void make_committed_txn_visible(const std::string& instance_id, int64_t db_id, int64_t txn_id, |
Contributor
There was a problem hiding this comment.
warning: function 'make_committed_txn_visible' exceeds recommended size/complexity thresholds [readability-function-size]
void make_committed_txn_visible(const std::string& instance_id, int64_t db_id, int64_t txn_id,
^Additional context
cloud/src/meta-service/txn_lazy_committer.cpp:200: 82 lines including whitespace and comments (threshold 80)
void make_committed_txn_visible(const std::string& instance_id, int64_t db_id, int64_t txn_id,
^| DCHECK(txn_id > 0); | ||
| } | ||
|
|
||
| void TxnLazyCommitTask::commit() { |
Contributor
There was a problem hiding this comment.
warning: function 'commit' exceeds recommended size/complexity thresholds [readability-function-size]
void TxnLazyCommitTask::commit() {
^Additional context
cloud/src/meta-service/txn_lazy_committer.cpp:296: 138 lines including whitespace and comments (threshold 80)
void TxnLazyCommitTask::commit() {
^
TPC-H: Total hot run time: 38230 ms |
TPC-DS: Total hot run time: 192287 ms |
ClickBench: Total hot run time: 30.99 s |
gavinchou
reviewed
Aug 28, 2024
gavinchou
previously approved these changes
Aug 28, 2024
Contributor
Author
|
run buildall |
| DCHECK(txn_id > 0); | ||
| } | ||
|
|
||
| void TxnLazyCommitTask::commit() { |
Contributor
There was a problem hiding this comment.
warning: function 'commit' exceeds recommended size/complexity thresholds [readability-function-size]
void TxnLazyCommitTask::commit() {
^Additional context
cloud/src/meta-service/txn_lazy_committer.cpp:296: 139 lines including whitespace and comments (threshold 80)
void TxnLazyCommitTask::commit() {
^
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 38261 ms |
TPC-DS: Total hot run time: 195283 ms |
ClickBench: Total hot run time: 31.9 s |
Motivation:
* In cloud mode we use `foundationdb` for storing rowset meta, when we load data with many rowsets,
`commit_txn` will failed because `foundationdb` currently limits all transactions to be below 10 MB in size,
So we spilt `commit_txn` into several sub fdb txn to solve the problem
How:
`commit_txn` flow like this
sub txn 1:
1. update partition `VersionPB` with txn_id
2. submit async task to txn_lazy_committer
txn_lazy_committer:
sub txn 2:
convert tmp rowset meta per batch
sub txn 3:
convert tmp rowset meta per batch
......
sub txn n:
make txn visible and remove txn_id in `VersionPB`
`get_version` main flow like this:
1. `VersionPB` without txn_id, the same as before
2. `VersionPB` has txn_id field, it means this is a txn
need be advanced by txn_lazy_committer, so submit a
`txn_lazy_task` to make last txn visible and retry
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 38018 ms |
TPC-DS: Total hot run time: 192851 ms |
ClickBench: Total hot run time: 32.53 s |
gavinchou
approved these changes
Aug 28, 2024
w41ter
approved these changes
Aug 29, 2024
w41ter
approved these changes
Aug 29, 2024
Contributor
|
PR approved by at least one committer and no changes requested. |
1 similar comment
Contributor
|
PR approved by at least one committer and no changes requested. |
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation:
foundationdbfor storing rowset meta, when we load data with many rowsets,commit_txnwill failed becausefoundationdbcurrently limits all transactions to be below 10 MB in size, So we spiltcommit_txninto several sub fdb txn to solve the problemHow:
Now
commit_txnmain flow like thissub txn 1:
VersionPBwith txn_idtxn_lazy_committer:
sub txn 2:
convert tmp rowset meta per batch
sub txn 3:
convert tmp rowset meta per batch
......
sub txn n:
make txn visible and remove txn_id in
VersionPB