[MDEV-31111] ERRORS: Invalid default value for 'check_time', DDL_LOG: Got error 1062 ... for entry 1 of type 'rename table', 1146: Table 'mysql.spider_table_sts' doesn't exist, ER_TABLE_EXISTS_ERROR and Got error 12524 'unknown' from SPIDER on RENAME/DROP TABLE Created: 2023-04-22 Updated: 2023-11-28 |
|
| Status: | Confirmed |
| Project: | MariaDB Server |
| Component/s: | Storage Engine - Spider |
| Affects Version/s: | 10.6, 10.7, 10.8, 10.9, 10.11, 11.0, 11.1 |
| Fix Version/s: | 10.6, 10.11, 11.0, 11.1 |
| Type: | Bug | Priority: | Major |
| Reporter: | Roel Van de Paar | Assignee: | Yuchen Pei |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | race, regression, regression-10.6, sporadic, sql_mode | ||
| Issue Links: |
|
||||
| Description |
|
The bug is highly sporadic, and there are likely a number of extraneous clauses in the testcase which are not necessary to reproduce the bug. Currently repeating the testcase about 500-1000 times, using the CLI, with a fresh server start for each t/c loop, reproduces the issue.
Leads to:
Further attempts to improve the test setup (MTR vs CLI), testcase length and reproducibility still in progress. Currently observed in 11.1 only, but likely present in other versions. |
| Comments |
| Comment by Roel Van de Paar [ 2023-04-22 ] | ||||||||||||||||||||||
|
Thus far all attempts at reducing/improving the testcase further have failed. Given the (ftm deemed) requirement to restart mariadbd in between tests replays take several hours. Perhaps an rr trace would be a good next step. | ||||||||||||||||||||||
| Comment by Roel Van de Paar [ 2023-04-24 ] | ||||||||||||||||||||||
|
Found a shorter testcase, which is as good as "non-sporadic", and reduced the same. No Foreign Keys are needed.
This leads to, across versions and build types, all these:
The CLI testcase output (when the bug occured) is this:
Interestingly, with the following small testcase change (GLOBAL sql_mode to use NO_ZERO_DATE instead of traditional):
We get the same issue. Additionally, with the following small testcase change (GLOBAL sql_mode not using traditional):
We instead see:
The test.t2 does not exist error could be valid here, given the double attempt in ALTER. | ||||||||||||||||||||||
| Comment by Roel Van de Paar [ 2023-04-24 ] | ||||||||||||||||||||||
|
With the following testcase:
We produce these interesting errors:
This looks to be an additional bug. | ||||||||||||||||||||||
| Comment by Roel Van de Paar [ 2023-04-24 ] | ||||||||||||||||||||||
|
This MTR testcase:
Produces both the
Errors in the error log. | ||||||||||||||||||||||
| Comment by Roel Van de Paar [ 2023-04-24 ] | ||||||||||||||||||||||
|
This MTR testcase:
Produces the
Error output in MTR. Note that ER_NO_SUCH_TABLE is not valid in either testcase in this comment. Additionally there is:
And,
Produces:
With the same warnings on the DROP. |