Re: bcache and load average

Going through various forum messages and bug reports (for Debian: 
https://lists.debian.org/debian-kernel/2015/03/msg00060.html for Arch: 
https://bugs.archlinux.org/task/38843 for some other distro: 
https://groups.google.com/forum/#!msg/esos-users/NXp8tG7sVE8/QXZyPdZ2saIJ 
) it looks like the bcache_writeback kernel thread, being in 
uninterruptible sleep, keeps the load average at 1.0 (or maybe more) 
always. Could you please confirm this ?

Try the attached patches that I've been collecting over the past year or 
two. I do not believe they have been merged into mainline (BUT SOMEONE 
NEEDS TO).

I am not sure that these address the load bug, but if the load is being 
increased by a large amount of dmesg output caused by rcu traces, then the 
patches will help.

-Eric
 

 - this behaviour should be described in the bcache documentation
   because it feels to me (and many other) like a true gotcha. It's
   apparently completely undocumented. A "CAVEATS" section at the
   bottom of bcache.txt in the kernel Documentation explaining this
   would be nice, what do you think?

 - Is there any way around this? Some people seem to grow uneasy (maybe
   irrationnally) having a constant load on an otherwise unused system
   (I know that a sleeping thread actually does nothing, but many 
   system administrators can't wrap their head around this idea). 

I've tried some advice I've found on the web, like switching to
writethrough and "echo 0 > /sys/block/bcache0/bcache/writeback_running"
but to absolutely no effect.

Any advice and idea is welcome :)

-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	[off-list ref]
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html