How does Percona inMemory handle aggregations that require allowDiskUse

We are looking to test if Percona might work for a VERY intense read pressure application that constantly fetches data from a MongoDB source with only about 2GB of backing data, BUT it also does many pipeline aggregations to structure the data returned that requires the “allowDiskUse” flag to be true for our MongoDB version of the database. The hope is that the “disk use” for the aggregation is also done in memory provided that the size is sufficient.

Has there been any experience using this flag and the In-Memory Percona Engine?

Like any good contributor, come back to answer your own question.

I set up Percona In-Memory 5.0.22 on an EC2 and loaded the 4GB of working data and 7GB working memory.

I ran the suspect aggregation without {allowDiskUse:true} and it failed as expected in the $group pipeline. Ran it with the flag and the query completed and saw a spike to the local drive in iotop both during the aggregation and during fetch cursor for 330,000+ results

So it appears it will work and fall back to the local disk for these operations.