Questions about mergerfs' garbage collection

[STPKITT]

Hello,
while investigating strange noises (sounding like heads becoming active & immediately parking again one or two times) from otherwise perfectly working HDDs in an OpenMediaVault system I saw log entries like mergerfs[394]: running basic garbage collection in 15 minute intervals and they were coinciding with the noise coming from the harddrives seemingly only when roughly a minute or so afterwards OpenMediaVault is polling S.M.A.R.T. values, which is set to every 30 minutes.
This brought me to the following general questions about mergerfs' garbage collection:

1. What does it actually do - why does garbage emerge at all from using mergerfs?
2. How long does it take? Judging by the harddrive activity LED it's only 1, 2 seconds though if that's true there shouldn't be a "collision" between the garbage collection and S.M.A.R.T.-polling.
3. Is garbage collection vital or worded differently what would be the drawback to not do it?
4. Is there a way to disable it?
5. Is there a way to configure the interval?

Thanks for reading!
Greets!

[trapexit]

1. Because it needs to manage it's memory. Every file, every node in the filesystem, requires allocation of memory. If you scan the whole filesystem and then do nothing it would still have all that memory allocated. For low end system that would be wasteful. Before adding domain specific allocators memory fragmentation would lead to memory bloat over time and kill low end systems.
2. It's extremely fast. Certainly not seconds. If it did you'd notice because the whole system would freeze.
3. It would bloat memory usage as with any gc behavior.
4. No
5. No

4 and 5 are no because it should have no impact on the system. It's extremely simple code and doesn't impact the underlying filesystems at all. You can look at the code. It's just doing a very minor gc of the node allocator and the message buffer allocator. Nothing there touches filesystems at all. If debugging is enabled it will write some metrics out to /tmp.

[STPKITT]

Thanks very much for your quick response which fully answered all my questions!
So it's the S.M.A.R.T. polls which cause the noises, which only come from the Seagate Archiv drives while the HGST UltraStars aren't bothered. I've now configured OMV's S.M.A.R.T. polling differently to avoid the Seagates' moving their heads for no practical reason.
Greets!