On Wed, Aug 4, 2010 at 3:10 PM, Kashif <melbogia+o...@gmail.com> wrote:
> I am using 'zfs send' and 'zfs receive' on two systems to backup some of the 
> datasets but at times, it is kind of frequent actually, the process gets 
> stuck and goes into uninterruptible sleep and stays there until I kill it.
>
> This is where it things get more exciting, in that I call kill the 'zfs send' 
> process on the source side but I cannot kill the 'zfs receive' process on the 
> receiving end and I have to reboot that machine.
>
> I am sending one dataset at a time so it's not like I am overburdening the 
> systems by spawning multiple zfs sends. I am using snv_111b on x86. Is this a 
> known bug? If so, are there any workaround?

I don't know the specific bug that causes this but I can confirm it's
a common situation in b111. After you kill the `zfs receive`, run
`iostat -XCn 1` and verify that it's reading a lot. When that activity
ends, `zfs receive` will have terminated.

b126 seems to be immune to this and can accept many `zfs receive`
processes without any of them getting stuck like that. You'd better
try b134 though.

-- 
Giovanni Tirloni
gtirl...@sysdroid.com
_______________________________________________
sysadmin-discuss mailing list
sysadmin-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss

Reply via email to