The Kerberos keytab file is specified in the KRB5_KTNAME environment variable.

Could a setParameter or other config file setting "krb5KtName" be provided to allow this to be set?

Please refer the Case: 00803199

We need to update to two binding accounts in the config file to for LDAP authentication so that we can avoid down time while resetting the binding account password.

It would be nice to have UUID support for $convert and $toString functions - and maybe also having a new $toUuid function added.

We have documents with UUIDs stored as UUIDs, and others where they are stored as Strings - and need to $lookup a document with an UUID type _id from a document where that uuid is stored as a String. As far as I can tell that is currently not possible.

It would be great if configuration changes could be effected without needed to restart nodes.

For example, audit filter, enabling/disabling different security mechanisms.

This would be especially useful for Atlas and clients with large clusters where restarts will cause a performance deficit due to cold cache after node restarts.

Just as a time-series collection in MongoDB 5.0 abstracts the underlying implementation of the bucketing pattern, customers who implement a multi-tenant model through separate databases per tenant run into the issue of too many dhandles and would benefit from an abstraction of the implementation of a collection-with-discriminator-field - they would be able to meet their internal compliance requirement of separate databases per customer, and would have minimal changes to their code.

We have test the retry write in Atlas by using mongodb java driver in Mongodb 4.4 cluster. The error is similar to SERVER-53624(https://jira.mongodb.org/browse/SERVER-53624). The supporter respond the error will fix in Mongodb 5.0. We hope this error can also fixed in Mongodb 4.4

a couple of customer a shifting from MongoDB to other DBs like Redis or even PostreSQL as count is too slow on MongoDB when there are a lot of document to count even with use of indexes

Two different scenarios. One scenario is to have a DR cluster. 2 independant clusters that are one way synced. As of right now, a custom CDC solution (custom code + Kafka) is needed to achieve this.
Another is because of latency and the requirement for data to be stored in 2 different data center, 2 indepentant mongoDB clusters that are two way synced. As of right now, a custom CDC solution (custom code + Kafka) is needed to achieve this.

Per the Server Support Matrix https://docs.mongodb.com/manual/installation/ support for Ubuntu 20 is in MongoDB Server version 4.4+ but not 4.2.
We would like to see the currently supported MongoDB Server version 4.2 available on the Ubuntu 20.4 LTS distribution.

As of right now you need to use the shard key as part of the query to make a target query to a specific shard. Would like the ability to make targeted query to a specific shard without using the shard key as a part of the query.

Maybe one way of doing this is using index metadata to avoid scatter gather query and using that index meta data to instead make targeted queries on sharded clusters.

NoTableScan at the collection level instead of mongod level.

In the options for Mongo DB Log Settings, there is only
Max Percent of Disk
Total Number of files

A new option for
Number of Days to keep

Would be useful

Is there any plan to implement distributed transaction that involves more than one data stores (for e.g. an RDBMS and Mongo)? We have one such requirement and tried a simple POC by creating a simple class extending XAResource (MongoXAResource implements XAResource) and overriding below methods.

@Override
public void commit(Xid xid, boolean b) throws XAException {
clientSession.commitTransaction();
}

@Override
public void rollback(Xid xid) throws XAException {
clientSession.abortTransaction();
}

It appears to work but i think there is a lot more to do. Is there any plan to implement by MongoDB team?

An expression index is one where the value being indexed is the result of an expression, like lower casing a string.

http://en.wikipedia.org/wiki/Expression_index
http://www.postgresql.org/docs/8.1/static/indexes-expressional.html

we've already seen a full text search and it will be awesome if you manage to implement partial version :)

Change streams currently return only the new version of the documents.

If there were an option that could be set on the collection for the change streams to also return the old version of the document, it would simplify any business logic that requires the old version.

Add support for repairing a single collection instead of what is currently available to repair an entire database via db.repairDatabase()

The global latency stats Server Status section currently aggregates all the latency information known to the server.

We have a need for the same information except aggregated by collection.

The actual command: db.serverStatus().opLatencies.