Go-FUSE logs are ambiguous with multiple servers

[Kern--]

We use go-fuse in soci-snapshotter - a containerd snapshotter that enables lazy loading of unmodified container images. Containerd creates a separate mount for each layer in a container image so that containers that share layers can share the content. For our implementation, this means each container layer has it's own fuse.Server which are layered using OverlayFS.

I'd like to be able to use the go-fuse logs to help debug workloads where lazy loading is performing poorly. Specifically, I'd like to be able to use the go-fuse logs to reconstruct all of the filesystem operations and their timings (we're adding timing information to the default logger) to either fix our implementation or drive optimizations in the workload itself.

Unfortunately, the unique id in the logs is only unique to a specific server so we end up seeing duplicates.

I propose adding an optional serverID string to fuse.Server. When logging operations, if the ID is not the empty, then add it in parenthases (e.g.):

rx 1(1): ... 
tx 1(1): ...
rx 1(2): ...
tx 1(2): ...

I'm happy to send a PR if you think this is a decent idea.

[navytux]

@Kern--, thanks for bringing this up.

To me your case suggests that instead of adding support for particular use case of serverID, we better add ability to generally specify log.Logger for fuse.Server, so that it is both possible to a) control where logs are sinked, and b) to add a custom prefix. The messages would then look like e.g.

srv1: rx 1: ...
srv1: tx 1: ...
srv2: rx 1: ...
srv2: tx 2: ...

@hanwen, @rfjakob, do you think that makes sense?

Kirill

[Kern--]

Yeah, that seems like a better solution and it would definitely solve my use-case.

[hanwen]

fixed in https://github.com/hanwen/go-fuse/commit/7b4f97c3577b3b799babe8a6cf3eebe41dde7193

[navytux]

Thanks, @hanwen.