So perusing the aio code in preparation to disassocate it from the Squid
fs code, I found that there's a pipe being used as a notification/wakeup
mechanism.
Basically, thread completion writes a ! into the pipe; this wakes up the
main squid loop and calls a PF callback.
All the callback does is
On ons, 2008-06-11 at 01:18 +0800, Adrian Chadd wrote:
Henrik,
Whats incomplete about the Variant invalidation support in PURGE/CLR in
Squid-2.HEAD?
There are many common corner cases where the existing vary_id isn't
picked up properly. The worst being if an object goes from not having
Vary
On tis, 2008-06-10 at 12:40 +0300, Mikko Kettunen wrote:
How could I configure squid to buffer whole POST-request before sending
it to peer cache or originserver?
Good question. It isn't really designed to do this. POST request may be
almost unbounded in size, and some streaming over HTTP
On Wed, Jun 11, 2008, Henrik Nordstrom wrote:
On ons, 2008-06-11 at 17:47 +0800, Adrian Chadd wrote:
How would it work? The diskd queues are all sysv msgq's which can't be
polled as filedescriptors. What would wake up the main Squid process?
make diskd output something on stdout after
On ons, 2008-06-11 at 17:48 +0800, Adrian Chadd wrote:
Subject says it all - open, read and unlink are async, but write and close
aren't.
Why is at least write not async?
Simply because when testing on Linux they very rarely block, with the fs
buffers absorbing them fully as long as you don't
On Wed, Jun 11, 2008, Henrik Nordstrom wrote:
Oh, forgot to say that there is #define's in store_asyncufs.h tuning
this. But the non-default settings hasn't been tested in a very very
long time.. (2000 timeframe)
Yeah, I'm testing async write at the moment. But I'm not exactly testing
it with
On Wed, Jun 11, 2008, Henrik Nordstrom wrote:
On ons, 2008-06-11 at 17:50 +0800, Adrian Chadd wrote:
Are there any objections to removing this patch from Squid-2.HEAD and
documenting this discussion (and other Vary related stuff) on a wiki page?
No problem.
Ok. I'll commit a reversal of
On ons, 2008-06-11 at 20:01 +0800, Adrian Chadd wrote:
What I need to do is get this Sun box actually running Solaris and try to
get dtrace to tell me what time is spent in the main thread blocking on
disk writes.
On Solaris you defenitely want async writes and closes, as Solaris UFS
by
Just for the sake of consistency and preference for better
troubleshooting and testing, are there any CFLAGS that I should be using
during configure and compile of Squid 3HEAD? I was using the same
options I was using with 2.6, which were:
CFLAGS=-DNUMTHREADS=65 -march=nocona -03 -pipe
Argh!
Please start putting relevant details into:
http://wiki.squid-cache.org/Features/VaryProcessing
That way we can have a proper list of whats required.
Adrian
On Thu, Jun 12, 2008, Mark Nottingham wrote:
These issues may be relevant:
http://tools.ietf.org/wg/httpbis/trac/ticket/37
10 matches
Mail list logo