Details
-
Bug
-
Status: Open (View Workflow)
-
Critical
-
Resolution: Unresolved
-
11.4, 11.8, 12.0(EOL)
-
None
Description
Numerous MTR tests have sporadic failures on buildbot from a mariadb connection being killed (disconnected?). A couple examples:
rpl.rpl_row_create_table can fail with
CURRENT_TEST: rpl.rpl_row_create_table
|
ERROR 2026 (HY000): TLS/SSL error: wrong version number
|
Error closing MYSQL pipe: at /home/buildbot/amd64-ubuntu-2204-debug-ps/build/mysql-test/var/9/tmpkrrPbn line 8.
|
mysqltest: In included file "./include/write_result_to_file.inc":
|
included from ./include/show_events.inc at line 91:
|
included from ./include/show_binlog_events.inc at line 35:
|
included from /home/buildbot/amd64-ubuntu-2204-debug-ps/build/mysql-test/suite/rpl/t/rpl_row_create_table.test at line 99:
|
At line 65: command "perl" failed with error: 1 my_errno: 0 errno: 0
|
And rpl.rpl_gtid_mdev4474 can fail with
CURRENT_TEST: rpl.rpl_gtid_mdev4474
|
ERROR 2026 (HY000): TLS/SSL error: Broken pipe (32)
|
Error closing MYSQL pipe: at /data/buildbot/workers/prod/amd64-freebsd-14/build/mysql-test/var/16/tmph37h8w line 8.
|
mysqltest: In included file "./include/write_result_to_file.inc":
|
included from ./include/show_events.inc at line 91:
|
included from ./include/show_binlog_events.inc at line 35:
|
included from /data/buildbot/workers/prod/amd64-freebsd-14/build/mysql-test/suite/rpl/t/rpl_gtid_mdev4474.test at line 49:
|
At line 65: command "perl" failed with error: 1 my_errno: 0 errno: 0
|
Both of these examples use write_result_to_file.inc which uses the mariadb command-line tool to execute a SHOW BINLOG EVENTS query. Though note, not all failures are limited to this use case of SHOW BINLOG EVENTS, but a fair number of them are, which makes me wonder if that command has a bug and causes mariadb cli to die/crash.
The relevant code in this use case in write_result_to_file.inc is in a perl script:
perl;
|
...
|
open MYSQL, "| $mysql > $outfile" or die "Failed to open pipe to client on server '$server_number': $!";
|
print MYSQL $stmt, ';' or die "Error printing statement to MYSQL pipe: $!";
|
close MYSQL or die "Error closing MYSQL pipe: $!";
|
EOF
|
Where the tests fail on the close line. These errors can be reproduced by providing write_result_to_file.inc with some long running command (e.g. set debug_sync="now wait_for nosig") and killing it while the helper script file is executing it.
So then the question is – why are these mariadb connections randomly dying during tests?
Attachments
Issue Links
- is part of
-
MDEV-36647 No red leaves in the forest
-
- Open
-