On lör, 2007-09-01 at 00:20 +1200, Amos Jeffries wrote:
I'm now getting compile problems.
Under the new state does USE_POLL or HAVE_POLL have preference over
runtime code? (src/structs.h:1567 versus src/stat.cc:993)
USE_POLL means configure has selected the poll select loop.
HAVE_POLL means
On tor, 2007-08-30 at 13:58 +1200, Amos Jeffries wrote:
On the side of this. I read through the PRE7 release notes and noticed
that while the squid.conf changes are covered there is no comparable list
of ./configure changes.
Right.. any takers on writing that?
Regards
Henrik
signature.asc
On tor, 2007-08-30 at 12:38 +0200, Henrik Nordstrom wrote:
On tor, 2007-08-30 at 13:58 +1200, Amos Jeffries wrote:
On the side of this. I read through the PRE7 release notes and noticed
that while the squid.conf changes are covered there is no comparable list
of ./configure changes.
On ons, 2007-08-29 at 00:34 +1200, Amos Jeffries wrote:
After todays changes I am now getting (with --enable-epoll)
checking for epoll_ctl in -lepoll... no
checking for epoll_ctl... yes
checking if epoll works... yes
Error - no epoll support found
Try running 'sh ./scripts/get_epoll
On ons, 2007-08-29 at 00:34 +1200, Amos Jeffries wrote:
After todays changes I am now getting (with --enable-epoll)
checking for epoll_ctl in -lepoll... no
checking for epoll_ctl... yes
checking if epoll works... yes
Error - no epoll support found
Try running 'sh ./scripts/get_epoll
After todays changes I am now getting (with --enable-epoll)
checking for epoll_ctl in -lepoll... no
checking for epoll_ctl... yes
checking if epoll works... yes
Error - no epoll support found
Try running 'sh ./scripts/get_epoll-lib.sh'
then run configure again
The script mentions does
lör 2006-08-12 klockan 09:51 +1000 skrev Robert Collins:
That is - isn't the for loop completely irrelevant ?
It is.. just a matter of programming style (complex if condition, vs
smaller conditions with break).
Regards
Henrik
signature.asc
Description: Detta är en digitalt signerad
lör 2006-08-12 klockan 11:22 +0200 skrev Henrik Nordstrom:
lör 2006-08-12 klockan 09:51 +1000 skrev Robert Collins:
That is - isn't the for loop completely irrelevant ?
It is.. just a matter of programming style (complex if condition, vs
smaller conditions with break).
This said I agree
for (;;) {
num = epoll_wait(kdpfd, pevents, SQUID_MAXFD, msec);
statCounter.select_loops++;
if (num = 0)
break;
if (ignoreErrno(errno))
break;
getCurrentTime();
PROF_stop(comm_check_incoming);
return
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Il giorno 21/lug/06, alle ore 23:31, Henrik Nordstrom ha scritto:
tis 2006-07-11 klockan 14:34 +0200 skrev Luigi Gangitano:
During discussions on the debian-devel mailing list, it was proposed
to enable a fallback if epoll() is not available
tis 2006-07-11 klockan 14:34 +0200 skrev Luigi Gangitano:
I've just packaged squid-2.6.STABLE1 and squid-3.0.PRE4 for debian,
enabling epoll() support at build time.
This obviously makes squid fail with kernels older than 2.6.x which
are still supported by debian.
During discussions
Hi all,
I've just packaged squid-2.6.STABLE1 and squid-3.0.PRE4 for debian,
enabling epoll() support at build time.
This obviously makes squid fail with kernels older than 2.6.x which
are still supported by debian.
During discussions on the debian-devel mailing list, it was proposed
ons 2006-04-05 klockan 06:39 -0700 skrev john allspaw:
ok. thanks a lot for your reply Steven.
taking a quick look just now, I am seeing 91 of those messages from the last
minute.
some more questions/info:
1. if it turns out that they are indeed harmless, (and not indicative of a
/var/tmp/build.14768/squid-2.5.STABLE9/src/comm_select.c
/var/tmp/build.14768/squid-2.5.STABLE9/src/comm_select.c: In function
`commEpollBackon':
/var/tmp/build.14768/squid-2.5.STABLE9/src/comm_select.c:1201: warning: unused
variable `ev'
-- Forwarded message --
Date: Wed, 13 Apr 2005 12:15:40 +0200
From: [Windows-1250] Pawe Staszewski [EMAIL PROTECTED]
To: [EMAIL PROTECTED], squid-users@squid-cache.org
Subject: Re: Odp: Re: [squid-users] how to apply epoll-2_5 patch to
squid2.5-stable9
Ok :)
Some comparisions
On Wed, 30 Mar 2005, Steven Wilton wrote:
I would have thought that if I've manually sync'd the remaining changes
between the s2_5 branch and the epoll-2_5 branch it would now pick up
any new changes in the s2_5 branch and only try and merge them.
If cvsmerge is happy then you are fine.
You only
On Tue, 15 Mar 2005, Steven Wilton wrote:
Yeah, if the interest is there, I would be happy to publish it.
There is interest.
I will have a quick look at the patches, and remove some unnecessary
code.
Ok.
Would it be better to try and make it co-exist with the select() and
poll() code before
user-mode, so --enable-cpu-profiling/-pg results should
be meaningful.
Attachment contains cpu_profiling for 160 req / sec of squid-3.0 with epoll().
From these results, it seems that you have reached the maximum request
rate possible (bounded by CPU usage). Hard to tell actually, since
was curious about throughput on epoll vs
poll/select. Never mind.
008.75| Connection.cc:485: error: 1/1 (s110) Connection timed out
008.75| error: raw write after connect failed
after these req / sec configuration setting.
Try enabling SYN cookies, and running squid with ulimit -HSn
?
---
squid 3.0 without epoll():
Squid Cache: Version 3.0-PRE3
configure options: '--prefix=/usr/local/squid3pre' '--with-aufs-threads=32'
'--with-descriptors=32768' '--with-pthreads'
'--enable-storeio=null,ufs,aufs' '--enable-debug-cbdata'
cache_mem 200MB
cache_dir null /dev/null
size : 512
I like to have your review on this. can we get more req / sec
satisfaction on this setup?
---
squid 3.0 without epoll():
Squid Cache: Version 3.0-PRE3
configure options: '--prefix=/usr/local/squid3pre' '--with-aufs-threads=32'
'--with-descriptors
hai david,
thanks for your reply.
i have tested with squid + epoll() using null fs. it is consuming 90% CPU
at peak rate ( 180 req / sec ). system cpu idleness is
in the range of 0.
Are you have any benchmarking results to epoll() with squid there?
how to integrate shared
65 35 0 0
Looks like you are running with (the same?) CPU bottleneck.
epoll's advantage is that CPU usage does not grow with the number of
idle TCP connections. If the number of concurrent connections is large,
and there are no idle connections, epoll should only give a small
increase
On Wed, 3 Nov 2004, Muthukumar wrote:
There has not been any benchmark on Squid-3 + epoll in a long time. The performance of this is not known.
How many requests being generated by squid + poll() / squid + select() on 32 bit hardwares?
As I said there has not been any benchmarks of Squid-3
is ufs.
On the same server, squid with select/poll did starve CPU, and the number of open fds
never bot bigger than 1600. When I started using epoll, the traffic and the number of
concurrent connections did increase. I recommend epoll (at leas on Linux).
Regards,
G
On Wed, 3 Nov 2004 08:46:54
I've been using squid3 with epoll support for a couple of months.
In my case, squid with poll/select did consume up to 100% CPU. With epoll, CPU usage
dropped to less than 10%.
Long term average max CPU usage:
http://webs.uolsinectis.com.ar/garana/x/cpu.4.png
With epoll, CPU usage over
Hello Henrik,
Thanks again.
Is there anyone benchmarked squid+epoll() on polygraph? How may I expect requests
satisfaction limit on Linux host
2.6.5-1.358 #1 i686 athlon i386 GNU/Linux platform?
There has not been any benchmark on Squid-3 + epoll in a long time. The performance
Hai Gonzalo,
I've been using squid3 with epoll support for a couple of months.
In my case, squid with poll/select did consume up to 100% CPU. With epoll, CPU
usage dropped to less than 10%.
It seems to be great. How many requests are being generated per second?
Are you using squid-3.0-pre3
Hello All,
I am preparing epoll() I/O method benchmarking with Polygraph (Polygraph 2.5.5)
with setup as,
squid + epoll():
Linux host 2.6.5-1.358 #1 i686 athlon i386 GNU/Linux
Squid Cache: Version 3.0-PRE3
configure options: '--prefix=/home/muthu/squidepoll
Adrian Chadd [EMAIL PROTECTED] writes:
On Thu, Feb 19, 2004, David Nicklay wrote:
Hi,
I second Henrik's suggestion. The epoll stuff actually works on 2.4
kernels with a patch, and almost none of those have the new epoll stuff
in glibc.
Also, it would be nice if someone could fix
On Fri, Feb 20, 2004, Kinkie wrote:
yup, this has to do with the order we're checking poll/select/epoll/kqueue.
Here's my configure.in patch, it seems to do the right thing.
Give it a shot?
If it's possible at all, I think it would be best if the polling method
detection could be done
On Thu, 19 Feb 2004, Adrian Chadd wrote:
On Fri, Feb 20, 2004, Robert Collins wrote:
just commit it.
ok.
Except for the first segment in the patch (-lepoll removed)
Regards
Henrik
On Thu, 19 Feb 2004, Adrian Chadd wrote:
Done. Can someone bootstrap it and commit it?
bootstrapping is done automatically before each nightly snapshot.
Regards
Henrik
On Fri, Feb 20, 2004, Henrik Nordstrom wrote:
On Thu, 19 Feb 2004, Adrian Chadd wrote:
On Fri, Feb 20, 2004, Robert Collins wrote:
just commit it.
ok.
Except for the first segment in the patch (-lepoll removed)
Crap, must've missed that. I'll put it back.
Adrian
Hi,
I second Henrik's suggestion. The epoll stuff actually works on 2.4
kernels with a patch, and almost none of those have the new epoll stuff
in glibc.
Also, it would be nice if someone could fix the autoconf scripts, so I
don't have to do:
--disable-select
--disable-poll
--enable-epoll
I
On Fri, 2004-02-20 at 14:19, Adrian Chadd wrote:
On Thu, Feb 19, 2004, David Nicklay wrote:
Hi,
I second Henrik's suggestion. The epoll stuff actually works on 2.4
kernels with a patch, and almost none of those have the new epoll stuff
in glibc.
Also, it would be nice
On Fri, Feb 20, 2004, Robert Collins wrote:
just commit it.
ok.
adrian
On Thu, Feb 19, 2004, Adrian Chadd wrote:
On Fri, Feb 20, 2004, Robert Collins wrote:
just commit it.
ok.
Done. Can someone bootstrap it and commit it? I've got a few
local hacks to get around the epoll linking that I'm thinking
of cleaner ways around..
adrian
On Fri, 2004-02-20 at 14:28, Adrian Chadd wrote:
On Thu, Feb 19, 2004, Adrian Chadd wrote:
On Fri, Feb 20, 2004, Robert Collins wrote:
just commit it.
ok.
Done. Can someone bootstrap it and commit it? I've got a few
local hacks to get around the epoll linking that I'm thinking
On Fri, Feb 20, 2004, Robert Collins wrote:
Done. Can someone bootstrap it and commit it? I've got a few
local hacks to get around the epoll linking that I'm thinking
of cleaner ways around..
I think Henrik has that cron'd.
Right. I thought something like that might be happening
On Tue, 17 Feb 2004, Adrian Chadd wrote:
Should we remove the -lepoll? Make configure test for its presence
if --enable-epoll is given?
configure test.
new kernel versions are adopted by the users much faster than new glibc
versions.
Regards
Henrik
On Wed, 12 Nov 2003, Gonzalo A. Arana wrote:
True, I've checked squid-3.0-PRE3-20031112 (today snapshot).
Btw, sory about coding style. I'll follow squid coding style on future
patches.
No problem. The reformatting is automatic by astyle and is not a burden
when accepting a patch.
The
+epoll)
Hi,
Ok, it seems that you are using wccp, am I right?
If so, please, could you test my patch again without wccp enabled?
Please disable as well (if enabled): external_acls, redirectors,
external auth, and use ufs as cache_dir.
I am curious. Is it only UDP (versus TCP) that you
On Wed, 12 Nov 2003, Gonzalo A. Arana wrote:
4) then, squid calls again commSetSelect(fd, COMM_SELECT_READ,
read_handler, ...). This is because a new request may be received from
client through same TCP connection
Yes. This is part of persistent connection management and also done as
part of
...
This was fixed in the patch which went into HEAD wasn't it?
True, I've checked squid-3.0-PRE3-20031112 (today snapshot).
Btw, sory about coding style. I'll follow squid coding style on future
patches.
Regards
Gonzalo
Hi,
Ok, it seems that you are using wccp, am I right?
If so, please, could you test my patch again without wccp enabled?
Please disable as well (if enabled): external_acls, redirectors,
external auth, and use ufs as cache_dir.
I am curious. Is it only UDP (versus TCP) that you are having
]
Subject: Re: new patch new ideas (squid+epoll)
Hi,
Ok, it seems that you are using wccp, am I right?
If so, please, could you test my patch again without wccp enabled?
Please disable as well (if enabled): external_acls, redirectors,
external auth, and use ufs as cache_dir.
I am
took some time to test this. Gonzalo's patch seems to
address a lot of outstanding issues with comm_epoll including:
- CPU usage problem w/ idle clients (both local and remote)
- epoll kernel table being out of sync with fde table
- missed epoll table updates (happens on my boxes
of:
Linux kernel (mine's 2.4.21 patched with
http://www.xmailserver.org/linux-patches/epoll-lt-2.4.21-0.18.diff)
Epoll-lib (mine's 0.10)
Pcl (mine's 1.2)
Reverting the patch and squid works fine again.
Sory to read that.
Thanks,
Reuben
I hope this helps,
Gonzalo
(both local and remote)
- epoll kernel table being out of sync with fde table
- missed epoll table updates (happens on my boxes but not Gonzalo's)
Could someone apply this to the squid-3.0 HEAD?
Kudos to Gonzalo.
Gonzalo Arana wrote:
Hi,
(yes, it's me again :-( ).
I have this new patch which
, David Nicklay wrote:
Hi,
I finally took some time to test this. Gonzalo's patch seems to address
a lot of outstanding issues with comm_epoll including:
- CPU usage problem w/ idle clients (both local and remote)
- epoll kernel table being out of sync with fde table
- missed epoll
problem w/ idle clients (both local and remote)
- epoll kernel table being out of sync with fde table
- missed epoll table updates (happens on my boxes but not Gonzalo's)
Could someone apply this to the squid-3.0 HEAD?
Kudos to Gonzalo.
Gonzalo Arana wrote:
Hi,
(yes, it's me again :-( ).
I
Henrik
On Thu, 6 Nov 2003, David Nicklay wrote:
Hi,
I finally took some time to test this. Gonzalo's patch seems to address
a lot of outstanding issues with comm_epoll including:
- CPU usage problem w/ idle clients (both local and remote)
- epoll kernel table being out of sync with fde
On Thu, 2003-10-16 at 03:54, David Nicklay wrote:
Hi,
I have been looking at this in a little more detail now. Currently, I
am using read_handler and write_handler as variables to make decisions
about what epoll is interested in inside of the kernel. Now I can see
Hi,
The early epoll API was virtual impossible to write for, because it did
not do an initial poll on new file descriptors you register an interest
for, so it could not know about notifications that happened before then.
I just about tore my hair out trying to code around that, and still
had
Hi,
On Mon, 2003-10-13 at 09:37, Robert Collins wrote:
On Sat, 2003-10-11 at 06:30, Gonzalo Arana wrote:
Hi,
(I'm back to squid-gzip task now).
I come up to this situation:
squid 3.0-PRE3-20031008 with epoll
kernel 2.4.21 patched with
http://www.xmailserver.org/linux-patches
On Sat, 2003-10-11 at 06:30, Gonzalo Arana wrote:
Hi,
(I'm back to squid-gzip task now).
I come up to this situation:
squid 3.0-PRE3-20031008 with epoll
kernel 2.4.21 patched with
http://www.xmailserver.org/linux-patches/epoll-lt-2.4.21-0.18.diff
When a client requests a very long
Hi,
(I'm back to squid-gzip task now).
I come up to this situation:
squid 3.0-PRE3-20031008 with epoll
kernel 2.4.21 patched with
http://www.xmailserver.org/linux-patches/epoll-lt-2.4.21-0.18.diff
When a client requests a very long object (such as a video), squid uses
100% of CPU
Polygraph: Polygraph-2.55
Query:
# Then another query that anybody tested the squid with epoll support on
IA64 more than 300 requests,
Please suggest your memory consumption,kernel version and kernel parameters
and kernel usage for testing.
# Anybody checked the squid-3.0
problem on epoll i/o squid on IA64
Date: 04 Sep 2003 16:01:19 +0530
Hello all ,
I am currently working in a development works in epoll I/O on squid for satisfying the
requests of 2000 in IA64.I have passed only 300 req/sec in IA64.My squid is using more
than 1.9 GB outof 2 GB.I have tried a lot
On Thu, Sep 04, 2003, Henrik Nordstrom wrote:
I am currently working in a development works in epoll I/O on squid for satisfying
the requests of 2000 in IA64.I have passed only 300 req/sec in IA64.My squid is
using more than 1.9 GB outof 2 GB.I have tried a lot in the squid to overcome
Hi,
I did:
./configure --enable-epoll --disable-poll --disable-kqueue
--disable-select
which should have done this, but the USE_EPOLL 0 is what shows up.
On Thu, 2003-06-12 at 15:26, Henrik Nordstrom wrote:
On Thursday 12 June 2003 18.10, David Nicklay wrote:
Could someone with more
62 matches
Mail list logo