the current status of nullfs, unionfs

[Marc G. Fournier]

On Thu, 10 Mar 2005, Andy Hilker wrote:

> Hi,
>
> You (Denis Shaposhnikov) wrote:
>>>>>>> "Kris" == Kris Kennaway <kris at freebsd.org> writes:
>>
>>  Kris> nullfs seems to work fine, unionfs is very fragile and easily
>>  Kris> exploded.
>>
>> nullfs is absolutely useless for jail's because TOO slow.
>
>
> What do you mean exactly, how do you benchmark this?

That's okay, my experiences are that nullfs is too fragile and easily 
exploded ... unionfs itself, though, isn't *rock solid*, but considering 
the loads that I throw at it on a daily basis, its alot more solid then 
the man pages indicate ...

One server I have running 80+ jails on, each with their own unionfs 
running on it, and the last crash was a result of:

panic: ufs_direnter : compact2

uptimes on our servers, all of which use unionfs extensively:

pluto# ruptime
jupiter       up  11+23:09,     0 users,  load  0.65,  0.78,  1.06
mars          up  12+01:06,     0 users,  load  0.53,  0.50,  0.55
neptune       up   1+21:42,     3 users,  load  2.03,  3.16,  3.10
pluto         up  12+00:54,     3 users,  load  1.33,  1.12,  1.34
venus         up  12+00:00,     0 users,  load  2.99, 10.73, 15.76

venus/neptune each run >80 jails ...

And the reason why the other 4 servers are only around 12 days right now 
is because we just moved the servers to a new rack ...

All servers run 4-STABLE, some a bit older, but a couple are less then a 
month old ...


----
Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
Email: scrappy at hub.org           Yahoo!: yscrappy              ICQ: 7615664