Konrad Rzepecki:
> W dniu 18.01.2012 13:03, Wietse Venema pisze:
> > Konrad Rzepecki:
> >> I have similar problem, and I suspect you may also hit it. In my case I
> >> send ~10k-20k mails through sendmail wrapper. My problem is caused by
> >> fsync done on EVERY queued mail. It take on my servers (4 core AMD @
> >> ext4 @ LVM @ RAID10 @ 8disk) ~ 0.2-0.3s for each one. Sometimes it's
> >> even 0.6s! So simple mail queuing for 15k recipients can take 1 hour or
> >> more...
> >
> > 0.6s for fsync is awful. You need to use a better file system.
>
> I'm afraid this is not only fs issue, but also all layer under it. LVM
> and RAID10 must give some time footprint. And in best case data are
> written to 2 disks and in worse, due to raid and lvm chunks, to all 8 I
> have.
>
> However I've not test which layer is responsible for that.
Average disk latencies are under 5ms nowadays, so having to wait
300ms for an fsync means that something is very wrong. I won't
speculate if this is due to Linux fsync sabotage (solution: use a
different OS), absence of battery-backed cache in the RAID (a must
for write-intensive loads), or something else.
Wietse
