I'm adding systems in earnest to my new BackupPC 4.1.2 installation, and I've encountered an interesting problem.
My current pool uses a decent number of inodes:
cpool:7864978
pc:11504955
For a total of 19,369,933 objects.
However, my NFS filesystem, which had 50M pre-allocated inodes, and a max of 80M inodes briefly ran out of inodes! I increased the max, and it finally settled at 88M inodes allocated, but currently reports ~20M used. A number of double checks confirms the current useage. I was never able to show that the file system actually had 80M+ inodes, making me think they were quickly allocated and then released before I could catch the state.
Since BackupPC is the only process pointed at this file system, the conclusion is that BackupPC is, as some point in the process, requesting far more inodes for a short time, than it is using.
It looks like, based on the numbers, that BackupPC is attempting to temporarily allocate a guess at the number of inodes needed to store the $Conf{FullKeepCnt} which in my case was [4, 0, 4, 0, 0, 2] or 10.
Can a developer comment on under what conditions BackupPC might be temporarily allocating a lot of extra inodes, and then quickly releasing them?
Thanks.
--
Time flies like an arrow, but fruit flies like a banana.
Ray Frush
