Add application limits for Jira imports
When we do a Jira import, we can generate a lot of cache keys. gitlab-com/gl-infra/scalability#419 (comment 381941081) shows that we had ~760,000 keys in the cache at one point while a large import was running.
In general, this is fine, and the Jira importer is using Redis in an acceptable manner, especially as these keys only exist during import. However, we might want to consider a limit:
- We want to set application limits in general.
- As this is in the Redis cache instance, it means that we're at risk of either evicting keys that we want to use in the cache, or having these keys themselves be evicted.
Proposal
Edited by Gabe Weaver