-
Notifications
You must be signed in to change notification settings - Fork 3.7k
[fix](move-memtable) abstract multi-streams to one logical stream #42039
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
|
run buildall |
|
clang-tidy review says "All clean, LGTM! 👍" |
TPC-H: Total hot run time: 41398 ms |
TPC-DS: Total hot run time: 191852 ms |
ClickBench: Total hot run time: 32.56 s |
b2b35e9 to
e6dc184
Compare
|
run buildall |
|
clang-tidy review says "All clean, LGTM! 👍" |
TPC-H: Total hot run time: 41331 ms |
TPC-DS: Total hot run time: 192063 ms |
ClickBench: Total hot run time: 31.82 s |
|
run buildall |
|
clang-tidy review says "All clean, LGTM! 👍" |
TPC-H: Total hot run time: 41230 ms |
|
TeamCity be ut coverage result: |
TPC-DS: Total hot run time: 191416 ms |
ClickBench: Total hot run time: 32.61 s |
|
clang-tidy review says "All clean, LGTM! 👍" |
3815d06 to
8bbd87e
Compare
8bbd87e to
8667d71
Compare
|
clang-tidy review says "All clean, LGTM! 👍" |
TPC-H: Total hot run time: 41229 ms |
TPC-DS: Total hot run time: 192112 ms |
ClickBench: Total hot run time: 32.61 s |
liaoxin01
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
|
PR approved by at least one committer and no changes requested. |
|
PR approved by anyone and no changes requested. |
sollhui
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
…ache#42039) ## Proposed changes Currently, an upstream BE (sink_v2) will open multiple streams to a downstream BE (load_stream). If any of the streams fails, the use_cnt on the downstream BE will be messed up. The load_stream will not report any success tablets to the sink_v2 since in its view there are still unfinished streams. So fault tolerance when open streams is not meaningful in practical, and may cause data lost. i.e. Upstream think there is still working streams to transfer data, but downstream does not report any commit info. This PR removes fault tolerance when open multiple streams to the same backend. If any of the open fails, the upstream sink_v2 should mark the downstream BE as failed replicas.
…ache#42039) Currently, an upstream BE (sink_v2) will open multiple streams to a downstream BE (load_stream). If any of the streams fails, the use_cnt on the downstream BE will be messed up. The load_stream will not report any success tablets to the sink_v2 since in its view there are still unfinished streams. So fault tolerance when open streams is not meaningful in practical, and may cause data lost. i.e. Upstream think there is still working streams to transfer data, but downstream does not report any commit info. This PR removes fault tolerance when open multiple streams to the same backend. If any of the open fails, the upstream sink_v2 should mark the downstream BE as failed replicas.
Proposed changes
Currently, an upstream BE (sink_v2) will open multiple streams to a downstream BE (load_stream).
If any of the streams fails, the use_cnt on the downstream BE will be messed up.
The load_stream will not report any success tablets to the sink_v2 since in its view there are still unfinished streams.
So fault tolerance when open streams is not meaningful in practical, and may cause data lost.
i.e. Upstream think there is still working streams to transfer data, but downstream does not report any commit info.
This PR removes fault tolerance when open multiple streams to the same backend.
If any of the open fails, the upstream sink_v2 should mark the downstream BE as failed replicas.