kstat troubles

[sgheeren]

Emmanuel,

as you are probably aware there are a number of problems reported having to do with kstat not mounting:

• (A) in general kstat mount failing will lead to severe performance degradation and large number of logged messages saying

Jun 14 01:03:39 shadow zfs-fuse: kstat: fuse_mount error - trying to umount

• (B) kstat not mounting, prevents any filesystems from mounting (IO error). running zfs-fuse --no-kstat-mount removes all the symptoms (issue #60)
• (C) zfs umount -a borking with 'invalid dataset name' (issue #51) with mtab containing inexplicable
       dev/fuse /zfs-kstat fuse rw,nosuid,nodev 0 0
  instead of or besides
       kstat on /zfs-kstat type fuse (rw,nosuid,nodev,allow_other)
  

Do you have any thoughts on this?

I myself had reported the loss of performance after creating a broken daemon that couldn't do any fuse mounts (in a rejected issue #46

Should we (perhaps while we can) vote for a default config with no-kstat-mount?

[Emmanuel Anne]

The good question here is : how can this mount fail ???
(sorry I didn't look for it in the issues).
If /zfs-kstat doesn't exist it's created.
If the daemon crashed and zfs-kstat was not cleanly unmounted, it's detected and forced to unmount so that the mount won't fail.
So how can it fail anyway ?

The performance loss is because the mount can't be done when the daemon starts, it needs to be done once fuse is initialised and it can't know when it happens, so it tries this regularly if the mount has failed (in normal conditions, that's 2 or 3 tries, but if it fails all the time I guess it can slow things down).

Anyway you can disable it by default if you wish, it's really not a big issue, the command line option to disable it is here for that...

2010/6/16 sgheeren <sghe...@hotmail.com>

--
To post to this group, send email to zfs-...@googlegroups.com
To visit our Web site, click on http://zfs-fuse.net/

--
zfs-fuse git repository : http://rainemu.swishparty.co.uk/cgi-bin/gitweb.cgi?p=zfs;a=summary

[sgheeren]

On 06/17/2010 12:13 AM, Emmanuel Anne wrote:

The good question here is : how can this mount fail ???
(sorry I didn't look for it in the issues).
If /zfs-kstat doesn't exist it's created.
If the daemon crashed and zfs-kstat was not cleanly unmounted, it's detected and forced to unmount so that the mount won't fail.
So how can it fail anyway ?

The performance loss is because the mount can't be done when the daemon starts, it needs to be done once fuse is initialised and it can't know when it happens, so it tries this regularly if the mount has failed (in normal conditions, that's 2 or 3 tries, but if it fails all the time I guess it can slow things down).

Anyway you can disable it by default if you wish, it's really not a big issue, the command line option to disable it is here for that...

Yeah I guess so. I was hoping you could knit together the symptoms and come up with some kind of hypothesis/analysis.

For 'invalid dataset' I expect something is looking at mtab and not filtering kstat correctly all the time
Do you know when (why) the mtab line for kstat would show dev/fuse instead of kstat for the mounted object?
For nothing at all being mounted anymore (the most serious condition, only reported once, #60) I suspect broken error handling...

But how exactly... and why only on some machines...