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
Fix memory accounting via HTTP interface #11840
Fix memory accounting via HTTP interface #11840
Conversation
function perf_test() { time yes '127.1:8123/?wait_end_of_query=1' | head -n10000 | xargs -P10000 curl -s -d 'select 1' | grep -x -c 1 } function server() { local limit=$1 && shift clickhouse-server "$@" -- --max_server_memory_usage=$limit } perf_test: - before this patch with 1G limit: succeed only ~800 queries - after this patch with 1G limit: succeed ~8000 queries - before this patch with 10G limit: succeed only ~3000 queries - after this patch with 10G limit: succeed ~10000 queries Fixes: ClickHouse#11153
Broken in master @blinkov |
Something wrong with |
But the way, this test is also flaky but it fails only occasionally. |
I don't see the failure, but it may require some tuning |
e183f5a
to
1b1b940
Compare
w/o the patch the test fails with, since it succeed only 512 queries.
1b1b940
to
9744c99
Compare
@alexey-milovidov Ok, found, and fixed and now only 01091_num_threads fails. |
We have to ask @KochetovNicolai for help or maybe you can investigate? |
Actually there were two issues: - missing \n, so it picked the wrong query anway - no unique identifier, hence it may works incorrectly (and actually, event clickhouse-test script executes 'SELECT 1' query)
Interesting, fixed - 267a6c8 P.S. Hm, I didn't look that it does not fails on upstream, egh. |
Now tests passed |
…ceff19401bbc5b99b1716934b2827 Cherry pick #11840 to 20.5: Fix memory accounting via HTTP interface
Fix memory accounting via HTTP interface (cherry picked from commit b447508)
Changelog category (leave one):
Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
Fix memory accounting via HTTP interface (can be significant with
wait_end_of_query=1
)Fixes: #11153
Similar to: #11022