Compiling for COSS (was Re: squid-2.4stable2 ?)

Thanks Henrik,

This solved the SWAPFAIL_MISSes I was seeing (I guess the aio_read that
got pulled in from the aufs code when I configured with aufs was working
but not really as a substitute for POSIX aio_read).

Adrian, nevermind on the SWAPFAIL_MISSes and on the performance problems.

It actually appears to be pretty zoomy now at 70 req/sec, and with near
ideal hit ratio so far. I'm going to give it a good workout today and
tonight and will present findings later. I think you might just be on
to something here.

I will also look into SGI's kaio. I seem to recall converting from
POSIX to kaio isn't too big of a job. If it isn't I'll try a
linux_async_io.c (or perhaps just add some #ifdefs in the async_io.c
file, since there is only the three wrapper functions that call aio_xxx).

Is it safe to use multiple disks and cossdirs with the current COSS?

Henrik Nordstrom wrote:

> Joe Cooper wrote:
>
>
>>Interestingly, if I don't enable aufs/pthreads during configure I also
>>get an error for aio_read--but not when I do. I'm confused, and will
>>have to tackle this tomorrow, I reckon. I'm guessing aufs does
>>something that makes aio_read 'findable' but not the other two. Don't
>>ask me to explain how/why.
>>
>
> DON'T enable aufs if you are building a Squid using POSIX AIO. The function
> names used by aufs conflicts with POSIX AIO.
>
> Should have fixed this long ago. Now done in HEAD.

                                   --
                      Joe Cooper <joe@swelltech.com>
                  Affordable Web Caching Proxy Appliances
                         http://www.swelltech.com
Received on Mon Aug 20 2001 - 12:43:23 MDT