Hm. Sorry about the delay, I wanted to make sure that this was really
working before sharing the good news :-p
So the strangest thing happened. By doing these exact steps:
1) Disable ZFS shares
2) Enable /tmp ro share
3) Reboot
4) Enable ZFS shares
5) Remove /tmp share
everything now seems to work. I'm not sure exactly how/why, but it
fixed itself.
Rebooting took over twelve hours. It had to time out on something
while unmounting every zfs filesystem, which was painfully slow.
But it works now, and reboots are down to manageable times.
Thanks for all your help,
Steven
On Apr 10, 2009, at 12:45 PM, Tom Haynes wrote:
> Steven,
>
> I wrote a simple little script to create 3000 zfs filesystems and
> then to
> share the common ancestor:
>
> [th199096 at jhereg ~/scripts]> more tasty.pl
> #!/usr/bin/perl
>
> print "zfs create tank/home\n";
>
> for ($i = 0; $i < 3000; $i++) {
> print "zfs create tank/home/u$i\n";
> }
>
> print "zfs set sharenfs=rw tank/home\n";
>
> And then I executed the output:
>
> [root at jhereg ~]> uname -a
> SunOS jhereg 5.11 snv_111 i86pc i386 i86pc
> [root at jhereg ~]> zfs list | wc -l
> 3004
> [root at jhereg ~]> share | wc -l
> 3002
> [root at jhereg ~]>
>
> No clue as to how long it took - I went out to lunch.
>
> nfsd was not running before I started.
>
>
> So I don't think it is the number of shares and your other
> experiments have supported this.
>
> I can bump down to snv_101b, but I'm not sure that is the problem.
>
> Some questions:
>
> 1) svcs -xv only shows that nfsd is not running, right?
>
> 2) You have rebooted at some point? We all know that makes things
> work, right? :->
>
> 3) Did you change any other properties on the zfs data sets?
>
> 4) Did you change any nfs settings?
> /etc/default/nfs
>
> 5) When you tried Rob's 1 share, are you sure that all of the zfs
> sharenfs properties
> are cleared?
>
> ---
>
> I did a bug search and I saw one where secure-by-default caused nfsd
> not to start. But
> that was not reproducible.
>
> Something to check though.
>
> Thanks,
> Tom
