One loses storage efficiency with that kind of partitioning. Elastic's watermarks is supposed to kind of do that without losing that efficiency due to external fragmentation.
Not sure If we are talking about the same concept. In Linux the term partition is unambiguous but I think in windows they are most frequently called volumes. What I mean is to divide the space available between several virtual disks, which from the point of view of the applications are isolated drive units. I use the term partition as I guess what you get in the VPS is a single disk unit, which then could be partitioned, like a physical disk can be in any OS.
In that scenario, the Elastic wouldn't see the other ones, it would had its own disk.
Because each application has its own disk (volume), no application can take out space from another one, so it's a way of isolating the problems. For example, this is why in Linux servers there is always a partition for the boot, to ensure that even if all disks get full, the system can boot with basic capabilities. Or ff the watermark of the Elastic is reached, you know that isn't going to affect the database. Aligning the physical structure of the partitions (really virtual but physical from the point of view of the applications) with the operational needs is a rather common practice among sys admins. I haven't seen any server (I've only worked with Linux servers) where all important components didn't have their own disk (partition, volume). That way, each disk has a particular needs of backup, performance, etc. aligned with the actual need of the application. For instance, a database could benefit from a type of disk and certain tuning, so at any point you can add a different disk and move the entire database to the new one just changing the
mount point (this is Linux jargon). It also facilitates backups and alarms. For instance, if you have and alarm on the database partition, it's easier to interpret immediately than an alarm on a disk that contain all the applications. In many managed servers like VPSs you can have backups but they are from whole disks, hence again, if you have a partition for the database you can copy the whole unit which is far more easy than selecting a bunch of folders.
Finally, the reasoning behind leaving some space unassigned is simply that enlarging a disk partition is a trivial and risk free operation in any OS, whereas shrinking one is a rather costly and potentially troublesome one. So there is always a tendency to be conservative about the space given to each disk (partition) and then watch each one and give more space individually as needed. In the contrary, if you
I understand that in managing a single server with limited time and all other constraints and circumstances is very tempting to have a single disk for everything. It can work certainly, but it has a tendency to be easy at first and then difficult to maintain and grow later. Sys admins tend to focus on minimizing the difficulty of maintenance and invest a lot of time at the beginning.
Of course, this all comes from experience with Linux servers, with Windows I have no experience at all at this level, so perhaps the "tradition" is different there.
In the end the most important thing is to understand the consequences of the two approaches, single disk unit vs multiple ones, and commit to one of them.
Again, good luck!
