As we adapt Elasticsearch to larger and larger clusters (see e.g. #77466) I've noticed something of an upwards trend in issues related to high alias counts. I've particularly noticed cases where the user has many (thousands or more) aliases pointing to a single index, whereas the implementation in Elasticsearch is more targetted towards situations where there are relatively few aliases each of which points to a large number of indices.
I don't think we have any docs that give guidance about alias usage or sizing, nor do we have any safety limits on alias counts, and this makes it hard for users to design their system to use aliases in Elasticsearch appropriately.
As we adapt Elasticsearch to larger and larger clusters (see e.g. #77466) I've noticed something of an upwards trend in issues related to high alias counts. I've particularly noticed cases where the user has many (thousands or more) aliases pointing to a single index, whereas the implementation in Elasticsearch is more targetted towards situations where there are relatively few aliases each of which points to a large number of indices.
I don't think we have any docs that give guidance about alias usage or sizing, nor do we have any safety limits on alias counts, and this makes it hard for users to design their system to use aliases in Elasticsearch appropriately.