Core Machine
Nucleus/Rock
OS: Version 1.0 (build 227) stable
Server: Version 1.8 (build 790) stable
HW: Intel NUC10i7FNH, 32 G RAM, 1 TB SSD
Network Details
CAT 6 cabling, linksys/intel 100/1000 switches
NFS Server: pi4, Rasbian, 12 TB USB3 dis
Audio Devices
Irrelevant to problem: Benchmark DAC-3
Description of Issue
I have a large library. Roon successfully imported about 227,000 tracks. 267 tracks were listed in the roon “skipped files” list as corrupt. Another 372 tracks were listed as “i/o failure”. These appear to be separate issues.
- corrupt files: 224 of these are digitized 78 tracks: low bit rate CBR mp3s, mostly from archive.org. I have over 1000 more 78 tracks that were succesfully imported. When I checked the files roon lists as corrupt by playing them with foobar, directly off the server, they play fine, and using foobar properties/details, I can read the tags. I found no errors in the dozen or so tracks that roon reports as corrupt that I tested, either in playback or tags. One possibility is that roon does not handle low rate mp3 well: some of these are 40 kbps historic recordings.
- i/o failure: this is a more serious problem. The 372 tracks listed as i/o failure all appear fine (again checking by playing them off the fileserver using foobar). They tend to be mostly entire albums (but not entirely). It appears that the failures can be repeated (limited testing) with repeatable results, and that the failures are mostly (but not entirely) complete albums. If the problem were network or disk, I would not expect repeatable failures. The fileserver logs have no disk errors. I’m not sure how your rescan works, but I tried “touching” an album and tracks to change its date, then rescanning in roon, and the error appeared to repeat. (in my opinion user visibility and control over the scanning process is a roon weakness).
Another observation is that before migrating to my current ROCK setup, while testing the core on a windows laptop, I’m almost certain that the skipped files from the windows scan was a very small number, under 20 instead of nearly 700. Perhaps there is a difference in how the scan performs in the stripped down linux version, and how it works on windows.
Please advise, thanks.