Thanks. I was referring not to the queries that use temp tables for their execution, but the queries that select from INFORMATION_SCHEMA.TEMPORARY_TABLES or INFORMATION_SCHEMA.GLOBAL_TEMPORARY_TABLES. Does your parallel workload include such queries?
For Percona Server 5.5.25, about to be released shortly, we have fixed one bug that is slightly similar to yours but requires the aforementioned queries to trigger (bug 951588).
In case you don't query these I_S tables, it'd be great if you could narrow down the parallel workload to some specific queries that cause ALTER TABLE hang. In that case complete stacktraces for all threads would be helpful too.
Ivor -
Thanks. I was referring not to the queries that use temp tables for their execution, but the queries that select from INFORMATION_ SCHEMA. TEMPORARY_ TABLES or INFORMATION_ SCHEMA. GLOBAL_ TEMPORARY_ TABLES. Does your parallel workload include such queries?
For Percona Server 5.5.25, about to be released shortly, we have fixed one bug that is slightly similar to yours but requires the aforementioned queries to trigger (bug 951588).
In case you don't query these I_S tables, it'd be great if you could narrow down the parallel workload to some specific queries that cause ALTER TABLE hang. In that case complete stacktraces for all threads would be helpful too.