Hi, clint On May 05, clint--- via discuss wrote:
I suspect this is something more recent as this issue just showed up with 10.11.6 for us. Short of valgrind or dumping a (massive) core file, is there some other way to see current allocations in mariadb?
It's likely not MariaDB as such - see MEMORY_USED in the INFORMATION_SCHEMA.PROCESSLIST and in SHOW GLOBAL STATUS, or check PERFORMANCE_SCHEMA. In all those cases we've heard about, the memory growth was caused by memory fragmentation, that is the memory is neither allocated by MariaDB nor available to the OS. Switching to a different memory allocator, like jemalloc or tcmalloc, have reportedly helped some users. Also, try to disable transparent huge pages. If you'll search for "transparent huge pages databases" you'll find a lot of links like https://www.pingcap.com/blog/transparent-huge-pages-why-we-disable-it-for-da... https://www.mongodb.com/docs/manual/tutorial/transparent-huge-pages/ https://docs.oracle.com/en/database/oracle/oracle-database/18/ladbi/disablin... https://www.percona.com/blog/why-tokudb-hates-transparent-hugepages/ etc. Or upgrade to 10.11.7 where MariaDB does it for you automatically. https://jira.mariadb.org/browse/MDEV-33279 Regards, Sergei Chief Architect, MariaDB Server and security@mariadb.org