[zfs-discuss] Re: Significant "pauses" during zfs writes

[Michael Schuster - Sun Microsystems]

Roch wrote:
> Michael Schuster writes:
>  > IHAC who is using a very similar test (cp -pr /zpool1/Studio11
>  > /zpool1/Studio11.copy) and is seeing behaviour similar to what we've
>  > seen described here; BUT since he's using a single-CPU box (SunBlade
>  > 1500) and has a single disk in his pool, every time the CPU goes into
>  > "100%-mode", interactive usability is reduced to zero for 1-2
>  > seconds. This is of course undesirable.  
>  > 
>  > + Do we have any bug IDs or RFEs addressing this? 
>  > + Is there any tuning the customer can try (he's willing to do anything with the machine, apart from donating it to me :-)? (we've already tried "echo txg_time/W1 | mdb -kw", with no improvement - on the contrary, machine returned to "usable" state hardly at all). 
>  > + has customer (and I) missed a constraint about when/where to deploy zfs?
>  > 
>  > thx
>  > Michael
>  >  
>  >  
>  > This message posted from opensolaris.org
>  > _______________________________________________
>  > zfs-discuss mailing list
>  > zfs-discuss at opensolaris.org
>  > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
> 
> 
> 
> What is the desired behavior ?

a system that doesn't freeze every few seconds just because someone is busy 
on a zpool.

> The workload (cp with compression) overloads the single
> CPU. Response time is bound to be affected. 

I wasn't aware that compression was involved here. Does raidz imply 
compression? I thought not.

also note: my customer is NOT using compression.

> 
> I am not sure how we could get compression to run at a low
> priority level, holding up write syscalls.
> 
> I suggest to Buy a second CPU or turn off compression.

see above - compression is off (I checked :-)

Michael
-- 
Michael Schuster                  +49 89 46008-2974 / x62974
visit the online support center:  http://www.sun.com/osc/

Recursion, n.: see 'Recursion'