vertebrateresequencing / muxfys Goto Github PK

When CacheBase is on an nfs mount, created cache dirs get left behind due to the RemoveAll in remote.deleteCache() failing with errors like .nfs0000000000b681f500000fb3: device or resource busy.

This seems to be related to github.com/alexflint/go-filemutex not closing its files on Unlock().

However, if altered to close on Unlock, the error becomes the even stranger: "directory not empty" (the directory in question containing an .nfs0000... file). Doing a rm -fr afterwards still gives a "Device or resource busy" error on the .nfs file.

I could not duplicate the latter problem with a direct test on go-filemutex.

Immediately after the go code exits, rm -fr works fine to delete the cache dir.

[bwt_restore_sa] SA-BWT inconsistency: primary is not the same. Abort!

Happens when (via wr) we have a mount like Mounts:[{"Mount":"npg-repository","Targets":[{"Path":"npg-repository","CacheDir":"/tmp/mounts_cache"}]}. The first bwa command to run works, but a subsequent one using the same cache fails.