Code review comment for lp:~laurynas-biveinis/percona-server/merge-5.6.12

Revision history for this message
Alexey Kopytov (akopytov) wrote :

The fix for http://bugs.mysql.com/bug.php?id=69419 looks problematic. We seem to be deviating from upstream (with all implications for users and performance_schema autosizing) merely to make a few tests pass in specific OS configurations.

Aren't there any better (less invasive) solutions? I didn't investigate the problem in details, but perhaps tweaking our Jenkins slaves would fix it without changing the server behavior? Or, if table_open_cache and table_definition_cache get some specific fixed values once the OS open_files limit is below a certain threshold, could we allow both "below threshold" and "above threshold" values in test results, e.g. with "--replace_result 615 1400"? Or, as the comment in the report suggests, check the actual (possibly adjusted) value of open_files_limit and skip the relevant tests if its below the threshold?

review: Needs Information

« Back to merge proposal