To answer the question directly, I would say not to change your root partition (which, I'm assuming, contains all of your OS files - i.e. your kernel) file system to XFS. By all means, use XFS on your new disk system for your Mongodb data.
I'm not sure of your exact disk setup, but if you have multiple disks, I would recommend OS on one (set), swap/tmp on another, database software on another, and actual data on yet others - that way you don't have OS calls blocking db stuff and vice versa. This also helps with restoring in the (unlikely) event of a disk (set) failure.
From here, you are correct in saying that
When running MongoDB in production on Linux, you should use Linux
kernel version 2.6.36 or later, with either the XFS or EXT4
filesystem. If possible, use XFS as it generally performs better with
MongoDB.
With the WiredTiger storage engine, use of XFS is strongly recommended
to avoid performance issues that may occur when using EXT4 with
WiredTiger.
From here (generally a good site), you can see that xfs is good for large files:
XFS was developed by Silicon Graphics in 1994 for the SGI IRX
operating system, and was ported to Linux in 2001. It’s similar to
Ext4 in some ways, as it also uses delayed allocation to help with
file fragmentation and does not allow for mounted snapshots. It can be
enlarged, but not shrunk, on the fly. XFS has good performance when
dealing with large files, but has worse performance than other file
systems when dealing with many small files. It may be useful for
certain types of servers that primarily need to deal with large files.
Now, I'm not sure about Mongodb, but as a database system, it probably has large files :-) - however, the kernel itself is primarily developed on ext4 and it is the default for a reason.
Remember that a recommendation is just that: a recommendation for an average set of databases with an average set of requirements! You should, if possible, benchmark with your own data on your own machines (prod and test!).
Another thing I've noticed about database file system benchmarks is that there's normally not a huge difference between them for a given task - obviously some are better than others in a given set of circumstances, but it seems to me that if one team makes a "breakthrough", others aren't far behind!
So, I would keep my kernel on ext4 and use xfs for my mongodb storage (again, with the caveat about your own benchmarking, to the extent that this is possible).
To answer @NeilLunn 's point: well, yes (what I wrote is essentially an opinion), but I like to think it is a reasonably educated one! Basically, there are lies, damned lies, and statistics! And then, there are database benchmarks :-)
Anything anyone writes on this topic will contain an element of opinion - the thing which separates the wheat from the chaff is those which take information from reputable sites, and those which urge the user to benchmark for their own apps on their own machines with their own realistic data sets and usage patterns.
I believe that I have done that - I'm not going to gainsay Mongodb about whether their systems work work better with file system x, y or z, but I also know that gainsaying Linus Torvalds on which is the best file system for the kernel is also fraught with peril!
