Plancache filling up /var/lib/memsql and causing outage

Hi,

In one of our Class 1 application, plancache filled up nearly 100GB within 15 minutes on MA and at the same time occupying full /var/lib/memsql which was sized at 128GB. This resulted in a highly critical cluster outage.

I spoke to our SingleStore TAM. We would need a feature where we limit the plancache based on space occupied than based on the days to retain. Right now, only option is disk_plancache_expiration_minutes . We would need something like disk_plancache_retention_max_size.

I landed up in a bridge call on a weekend and a problem ticket has been assigned to me.

Thanks,
Vasanth

This is solved in 7.6 with

7.6 Release Notes

enable_compiled_images_eviction

and compiled_images_eviction_memory_limit_mb

and existing LRU plan cache eviction.

Thanks a lot Hanson for this feature added. Is this only in memory or applies to disk also?

.

This is for in-memory only. On-disk plan expiration logic is described here: