Dropped collections on 3.6.23 secondaries stuck pending

Ops and Admin Replication
#1

Hi,

Since upgrading to 3.6 (yes I know!) we have found that dropped collections remain in pending on secondaries.

We’re running a primary-secondary-arbiter setup. The primary immediately frees disk space for dropped collections. The secondary however keeps the dropped collections in pending, and doesn’t free disk space.

We’re working around the problem by periodically restarting mongod on the secondary as drop-pending collections are removed on shutdown.

Could this be a configuration problem? We’re mostly using default settings (WiredTiger with journal enabled).

Thanks!

Martin.

#2

data pruning is usually done as a background task, so they may sit there for some time.

Does this cause any issue to your server?

#3

We have some collections that hold daily data that are a few GB in size. A new collection is created each day and we keep the last x days in Mongo. We archive and drop the oldest collection daily. The secondary is therefore growing a few GB every day, until we manually force it to shutdown and restart.

#4

Can anyone help me to restore the collection that is in a pending state after the drop?
I accidentally deleted a collection. I run db.getCollectionNames({includePendingDrops: true}) and I get the list of the collections that are in a pending state. Is there any way to recover them?