Обсуждение: [COMMITTERS] pgsql: Fix confusion of max_parallel_workers mechanism followingcrash.
[COMMITTERS] pgsql: Fix confusion of max_parallel_workers mechanism followingcrash.
От
Robert Haas
Дата:
Fix confusion of max_parallel_workers mechanism following crash. Commit b460f5d6693103076dc554aa7cbb96e1e53074f9 failed to contemplate the possibilit that a parallel worker registered before a crash would be unregistered only after the crash; if that happened, we'd end up with parallel_terminate_count > parallel_register_count and the system would refuse to launch any more parallel workers. The easiest way to fix that seems to be to forget BGW_NEVER_RESTART workers in ResetBackgroundWorkerCrashTimes() rather than leaving them around to be cleaned up after the conclusion of the restart, so that they go away before rather than after shared memory is reset. To make sure that this fix is water-tight, don't allow parallel workers to be anything other than BGW_NEVER_RESTART, so that after recovering from a crash, 0 is guaranteed to be the correct starting value for parallel_register_count. The core code wouldn't do this anyway, but somebody might try to do it in extension code. Report by Thomas Vondra. Patch by me, reviewed by Kuntal Ghosh. Discussion: http://postgr.es/m/CAGz5QC+AVEVS+3rBKRq83AxkJLMZ1peMt4nnrQwczxOrmo3CNw@mail.gmail.com Branch ------ master Details ------- http://git.postgresql.org/pg/commitdiff/8ff518699f19dd0a5076f5090bac8400b8233f7f Modified Files -------------- src/backend/postmaster/bgworker.c | 48 ++++++++++++++++++++++++++++++++++----- 1 file changed, 42 insertions(+), 6 deletions(-)
On Tue, Apr 11, 2017 at 12:52 PM, Robert Haas <rhaas@postgresql.org> wrote: > Report by Thomas Vondra. Patch by me, reviewed by Kuntal Ghosh. Ah, nuts. Sorry for misspelling your name, Tomas. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company