Garbage collection

[Matteo Calorio]

Hello,

is there a way to force the emptying of some storage quickly?

We have a test installation on PBS on a VM with 8 CPU, 128GB RAM and a 100TB datastore on a SAN shared via NFS (probably not the more efficient solution).

Last two GC failed, so now we are reaching the share saturation:



The state of the VM seems to be non-critical and the "prunes" are performed correctly:



The same on PVE interface:



Also the SAN shows no problems.

This is the current status of GC:

2022-04-19T09:31:43+02:00: starting garbage collection on store san05
2022-04-19T09:31:43+02:00: Start GC phase1 (mark used chunks)
2022-04-19T11:36:25+02:00: marked 1% (77 of 7611 index files)
2022-04-19T13:28:50+02:00: marked 2% (153 of 7611 index files)
2022-04-19T16:44:47+02:00: marked 3% (229 of 7611 index files)
2022-04-19T18:54:39+02:00: marked 4% (305 of 7611 index files)
2022-04-19T21:31:26+02:00: marked 5% (381 of 7611 index files)
2022-04-19T23:20:44+02:00: marked 6% (457 of 7611 index files)
2022-04-20T02:09:04+02:00: marked 7% (533 of 7611 index files)
2022-04-20T13:25:40+02:00: marked 8% (609 of 7611 index files)
2022-04-20T21:45:15+02:00: marked 9% (685 of 7611 index files)

Another question is: if I understood correctly phase1 touches all chunks and phase2 remove all chunks with atime > 24h 5m. But... what if phase1 lasts more than 24h 5m?

Thanks,
Matteo

 

[dietmar]

Hello,

is there a way to force the emptying of some storage quickly?

You need to use a reasonable fast storage.

 

[Matteo Calorio]

Oh... so the reply to my whole message is "no, there is no way to force the emptying of some storage quickly"? :/

Another question: is it possible to keep index file locally (SSD) and chunks on NFS?

M.