Unnecessary error messages for non-errors about go-nfs HOT 8 CLOSED

oh ho, it's supposed to be there but is not. It's almost as though it came back from a readdir but the stat failed?

One thing that would be helpful is some way to enable per-request logging, since it is not always possible to use wireshark (I'm running nfs3 over ssh).

More as I learn it.

from go-nfs.

Also, I should know this, but what is 0x10431b490

from go-nfs.

ah, ok, it seems to be the kernel trying to do something? maybe set atime? still looking.

from go-nfs.

OK, the code for lookup I think could be changed. nfs_onlookup should be able to use Lstat, not Readdir, as readdir is a very inefficient way to do a lookup (just image a directory with 10k+ entries and repeated lookups ... this happens. Or 1M. This happens).

I'm surprised that billy has no lookup operation, but I think Lstat would serve. Thoughts?

from go-nfs.

I was actually thinking the same thing (in practice, directories are not usually large enough to make a difference, but it's a simple optimization to make).

from go-nfs.

I just discovered LOG_LEVEL, which removes the extra prints. It seems to default to LOG_ERROR? I think it ought to be a higher level. Although, in general, log messages of any kind in a Go package are frowned upon ...

from go-nfs.

• when using this library in a larger program, you can connect it up to whatever logging system is in use via SetLogger https://github.com/willscott/go-nfs/blob/master/log.go#L61
• I think you're right that it may be worthwhile to differentiate between recoverable/non-recoverable errors at the library versus at the protocol level.
• The inefficiency in onlookup is a good call. I'll file a separate issue to track that.

from go-nfs.

I'd suggest removing this print. especially for shells, which do a lot of stats for PATH resolution, I am getting a lot of prints.

This is a file system error, and there are always a lot of these, and they are not really NFS-level failures,

from go-nfs.