On Thu, Sep 07, 2006, Henrik Nordstrom wrote:
> m??n 2006-09-04 klockan 11:17 -0600 skrev Brett Glass:
> > Henrik:
> >
> > In what version of Squid was this instability introduced? Does it
> > only show up in certain versions of the operating system? We have
> > caches running FreeBSD 4.9 and Squid 2.5.STABLE5 (see the output of
> > "squid -v" below) which are absolutely rock solid under heavy
> > loads; they work so well that we have not upgraded them.
>
> It's always been there actually, for as long as diskd has existed. But
> it only occurs when diskd is being pushed to the limits, which
> indirectly means the faster Squid is the likelier the problem is seen..
And I've been testing squid-2.6 + aufs on FreeBSD 6.1 and it seems quite
stable and performing (ie, not blocking on IO like it would do under
FreeBSD 4.x.)
Brett, could you perhaps try running Squid-2.6 + kqueue + aufs under
FreeBSD 6.x ? You should notice a bit of a performance boost.
I don't have high-spec enough hardware to really push things anymore
but I've run polygraph over a freebsd 6.1+squid-2.6+kqueue+aufs cache
with 2 IDE disks maxed out at a couple of megabytes a second and
it was all rather stable.
Adrian
Received on Wed Sep 06 2006 - 19:11:38 MDT
This archive was generated by hypermail pre-2.1.9 : Sun Oct 01 2006 - 12:00:03 MDT