Closed github-actions[bot] closed 3 weeks ago
The TileDB-MariaDB build failed at the step that confirms the linking with during the tests. There was no error message, so unclear what exactly happened. I restarted itldd
The centralized nightlies job failed on Thursday (2024-08-15) in run 10397650797
It failed again during the tests. It just stalls. And unfortunately it erases the log for some reason. The upstream TileDB-MariaDB nightlies are passing.
It also failed in my test build against TileDB 2.26.0-rc0 (https://github.com/jdblischak/centralized-tiledb-nightlies/commit/7de75e4cffa49487c28cfc6a39ea8d2afdc80f11, build)
This time I copy-pasted the test output when it stalled:
Logging: /home/runner/work/centralized-tiledb-nightlies/centralized-tiledb-nightlies/mysql-test/mariadb-test-run.pl --suite=mytile --debug
VS config:
Path length (125) is longer than maximum supported length (108) and will be truncated at /usr/lib/x86_64-linux-gnu/perl-base/Socket.pm line 193.
Too long tmpdir path '/home/runner/work/centralized-tiledb-nightlies/centralized-tiledb-nightlies/builddir/mysql-test/var/tmp' creating a shorter one...
- using tmpdir: '/tmp/_adyXPSPbK'
vardir: /home/runner/work/centralized-tiledb-nightlies/centralized-tiledb-nightlies/builddir/mysql-test/var
Removing old var directory...
Creating var directory '/home/runner/work/centralized-tiledb-nightlies/centralized-tiledb-nightlies/builddir/mysql-test/var'...
Checking supported features...
MariaDB Version 11.0.2-MariaDB-debug
- SSL connections supported
- binaries are debug compiled
Using suites: mytile
Collecting tests...
Installing system database...
==============================================================================
TEST RESULT TIME (ms) or COMMENT
--------------------------------------------------------------------------
worker[1] Using MTR_BUILD_THREAD 300, with reserved ports 16000..16019
mytile.1st [ pass ] 16
mytile.2nd [ pass ] 239
mytile.aggregates [ pass ] 850
mytile.create_allow_subset_existing_array [ pass ] 113
mytile.csv [ pass ] 83
mytile.data_types [ pass ] 1382
mytile.datetime_pushdown [ pass ] 1083
mytile.datetimes [ pass ] 1427
mytile.dense_writes [ pass ] 72
mytile.discovery [ pass ] 232
mytile.duplicates [ pass ] 145
mytile.empty_records [ pass ] 37
mytile.encryption [ pass ] 52
mytile.enum [ pass ] 222
mytile.estimated_table_records [ pass ] 124
mytile.fill_in [ pass ] 155
mytile.filters [ pass ] 318
mytile.flushing_arrays [ pass ] 230
mytile.heterogeneus_dimensions [ pass ] 140
mytile.hilbert [ pass ] 1490
mytile.indexes [ pass ] 942
mytile.insert_into_select [ pass ] 149
mytile.join [ pass ] 187
mytile.metadata [ pass ] 89
mytile.mrr [ pass ] 283
mytile.mrr_datetime_dimensions [ pass ] 260
mytile.mrr_heterogeneus_dimensions [ pass ] 234
mytile.mrr_incomplete_queries [ pass ] 303
mytile.mrr_string_dim [ pass ] 470
mytile.mrr_triple_join [ pass ] 577338
mytile.multi_value_fixed_length [ pass ] 109
mytile.nullable_attributes [ pass ] 1715
mytile.open_at [ pass ] 141
mytile.orderby [ pass ] 74
mytile.primary_key [ pass ] 164
mytile.pushdown_ranges [ pass ] 334
mytile.pushdown_two_predicates [ pass ] 224
mytile.query_conditions [ pass ] 330
mytile.schema_evolution [ pass ] 196
mytile.spatial_index [ pass ] 177
mytile.string_dim [ pass ] 231
And the tests stalled when running against TileDB 2.25.0 (build)
@DimitrisStaratzis do you have any ideas why the nightly TileDB-MariaDB tests are stalling here (against libtiledb dev, 2.26.0-rc0, and 2.25.0) but not in the upstream nightly build?
The centralized nightlies job failed on Thursday (2024-08-15) in run 10413538970
This is very weird, I will check
Could be related to the fact that MariaDB runs in debug mode in your builds and we have introduced some performance tests like the mrr_triple_join. Could you try building MariaDB in release mode ?
Could be related to the fact that MariaDB runs in debug mode in your builds and we have introduced some performance tests like the mrr_triple_join.
@DimitrisStaratzis Good idea! That is a clear difference between this nightly build and the upstream one. Though I am surprised that this problem is only surfacing now, when the upstream nightly build was updated 6 months ago (blame, diff).
Could you try building MariaDB in release mode ?
Testing now in https://github.com/jdblischak/centralized-tiledb-nightlies/commit/b28137e948cd2c7b2e02f8f19f48866d206c5e1e 🤞
Confirmed! By applying the changes from https://github.com/TileDB-Inc/TileDB-MariaDB/pull/304 to use Release mode instead of Debug mode, the tests are now passing.
Great! It is weird why it stalls in debug mode though, I will investigate!
The centralized nightlies job failed on Wednesday (2024-08-14) in run 10397650797