Hi John,
On Thu, May 15, 2014 at 07:41:25AM +0200, John-Paul Bader wrote:
Hey,
good news! HA-Proxy version 1.5-dev25-a339395 2014/05/10 is running
since yesterday morning with a 1/4 of our traffic and without coredumps,
infinite loops or other problems. We have nbproc and kqueue enabled
On 05/07/2014 12:35 PM, Vincent Bernat wrote:
❦ 7 mai 2014 11:15 +0200, Willy Tarreau w...@1wt.eu :
haproxy does not include DTrace probes by any chance right? :)
No, and I have no idea how this works either. But if you feel like it
can provide some value and be done without too much
Hey,
good news! HA-Proxy version 1.5-dev25-a339395 2014/05/10 is running
since yesterday morning with a 1/4 of our traffic and without coredumps,
infinite loops or other problems. We have nbproc and kqueue enabled and
everything seems to behave fine.
In peak times we had 11000 sockets open
On 07 мая 2014 г., at 18:24, Emeric Brun eb...@exceliance.fr wrote:
Hi All,
I suspect FreeBSD to not support process shared mutex (supported in both
linux and solaris).
I've just made a patch to add errors check on mutex init, and to fallback on
SSL private session cache in error
On Sun, May 11, 2014 at 12:19:45AM +0400, Dmitry Sivachenko wrote:
On 07 ?? 2014 ??., at 18:24, Emeric Brun eb...@exceliance.fr wrote:
Hi All,
I suspect FreeBSD to not support process shared mutex (supported in both
linux and solaris).
I've just made a patch to add errors
Hey Willy,
I have just applied the patch and will run another test after lunch.
Since we're testing with live traffic I can't leave it unattended :)
Just out of curiosity, is this a bug that affects also Linux or is it
FreeBSD specific?
Kind regards, John
Willy Tarreau wrote:
Good news!
Hi John,
On Fri, May 09, 2014 at 11:54:56AM +0200, John-Paul Bader wrote:
Hey Willy,
I have just applied the patch and will run another test after lunch.
Since we're testing with live traffic I can't leave it unattended :)
Just out of curiosity, is this a bug that affects also Linux or
to discover a crash of haproxy on an ALOHA,
and we couldn't explain it, it was in relation with a high memory
usage with SSL. Now I know where it was, so we'll issue an update :-)
Thanks again for all your traces and tests!
Willy
Can the 100% cpu load initially reported along with the segfaults
Hi Lukas,
On Fri, May 09, 2014 at 12:40:11PM +0200, Lukas Tribus wrote:
Can the 100% cpu load initially reported along with the segfaults be
explained by the unsupported process shared mutex on FreeBSD, btw?
Absolutely. Some lists are updated within the lock. I let you imagine
what a double
Hey again,
so for the first time haproxy is running for more than 4 hours without a
coredump and without cpu spikes - so far so good. I'll observe it until
tomorrow morning and report again.
Kind regards,
John
Willy Tarreau wrote:
Hi John,
On Fri, May 09, 2014 at 11:54:56AM +0200,
On Fri, May 09, 2014 at 04:52:04PM +0200, John-Paul Bader wrote:
Hey again,
so for the first time haproxy is running for more than 4 hours without a
coredump and without cpu spikes - so far so good. I'll observe it until
tomorrow morning and report again.
Great, so we might have addressed
Hey,
so I have downloaded the haproxy-ss-Latest from the website and applied
your patches. I have compiled it with:
make TARGET=freebsd USE_PCRE=1 USE_OPENSSL=1 USE_ZLIB=1
It ran very good for 2 hours but then 6 out of 12 processes coredumped,
this time however in the haproxy code realm and
Maybe the full backtrace is more helpful:
(gdb) bt full
#0 kill_mini_session (s=0x804269c00) at src/session.c:299
level = 6
conn = (struct connection *) 0x0
err_msg = value optimized out
#1 0x00463928 in conn_session_complete (conn=0x8039f2a80) at
Hi Willy,
When it uses the private cache, I would also have to change the
configuration to allow ssl sessions over multiple http requests right?
No you don't need to change anymore, what Emeric's patch does is to
reimplement a hand-crafted spinlock mechanism.
Two slightly unrelated
Hi John,
On Thu, May 08, 2014 at 09:15:20AM +0200, John-Paul Bader wrote:
Hey,
so I have downloaded the haproxy-ss-Latest from the website and applied
your patches. I have compiled it with:
make TARGET=freebsd USE_PCRE=1 USE_OPENSSL=1 USE_ZLIB=1
It ran very good for 2 hours but then 6
Hey,
Sure thing! I've just put an nokqueue in the config and its running
again. Lets see :)
Kind regards,
John
Willy Tarreau wrote:
Hi John,
On Thu, May 08, 2014 at 09:15:20AM +0200, John-Paul Bader wrote:
Hey,
so I have downloaded the haproxy-ss-Latest from the website and applied
your
Ok after about roughly another 2 hours the next core dump happened -
this time with the nokqueue option set.
The coredump looks very similar and now crashes with ev_poll instead of
ev_kqueue:
(gdb) bt full
#0 kill_mini_session (s=0x80337f800) at src/session.c:299
level = 6
On Thu, May 08, 2014 at 12:47:06PM +0200, John-Paul Bader wrote:
Ok after about roughly another 2 hours the next core dump happened -
this time with the nokqueue option set.
The coredump looks very similar and now crashes with ev_poll instead of
ev_kqueue:
Indeed it's exactly the same.
On Thu, May 08, 2014 at 01:46:30PM +0200, Willy Tarreau wrote:
On Thu, May 08, 2014 at 12:47:06PM +0200, John-Paul Bader wrote:
Ok after about roughly another 2 hours the next core dump happened -
this time with the nokqueue option set.
The coredump looks very similar and now crashes
Ok - I will happily test more patches to pin those issues down :)
There is one thing where I'm not sure if its related. We have currently
set up the system with a rather small msl setting:
net.inet.tcp.msl=5000
Also we have fast_finwait2_recycle enabled so we don't have that many
TIME_WAIT
Good news!
I found it by reading the code :-)
And I could even reproduce it. The bug happens when running with a
handshake (typically SSL) and when an out of memory or a socket error
happens when calling setsockopt(TCP_NODELAY), which might possibly be
made more common by your TCP settings,
Hi Lukas,
On Thu, May 08, 2014 at 10:20:28AM +0200, Lukas Tribus wrote:
Hi Willy,
When it uses the private cache, I would also have to change the
configuration to allow ssl sessions over multiple http requests right?
No you don't need to change anymore, what Emeric's patch does is to
Hey Willy,
this morning I was running another test without kqueue but sadly with
the same result. Here is my test protocol:
Running fine with nokqueue for about an hour at about 20% CPU per
process, then sudden CPU spike on all processes up to 90%, I started
ktrace but meanwhile the CPU
Hi John-Paul,
On Wed, May 07, 2014 at 09:22:32AM +0200, John-Paul Bader wrote:
Hey Willy,
this morning I was running another test without kqueue but sadly with
the same result.
OK so let's rule out any possible kqueue issue there for now.
Here is my test protocol:
Running fine with
Willy Tarreau wrote:
It's very interesting, it contains a call to ssl_update_cache(). I didn't
know you were using SSL, but in multi-process mode we have the shared context
model to share the SSL sessions between processes.
Yes, sorry. In the initial email on this thread I posted our
On Wed, May 07, 2014 at 10:28:18AM +0200, John-Paul Bader wrote:
Willy Tarreau wrote:
It's very interesting, it contains a call to ssl_update_cache(). I didn't
know you were using SSL, but in multi-process mode we have the shared
context
model to share the SSL sessions between processes.
On 05/07/2014 11:15 AM, Willy Tarreau wrote:
On Wed, May 07, 2014 at 10:28:18AM +0200, John-Paul Bader wrote:
Willy Tarreau wrote:
It's very interesting, it contains a call to ssl_update_cache(). I didn't
know you were using SSL, but in multi-process mode we have the shared
context
model to
Hey Emeric,
I have just consulted the Readme of the haproxy source and it says in
the OpenSSL section:
»The BSD and OSX makefiles do not support build options for OpenSSL nor
zlib. Also, at least on OpenBSD, pthread_mutexattr_setpshared() does not
exist so the SSL session cache cannot be
❦ 7 mai 2014 11:15 +0200, Willy Tarreau w...@1wt.eu :
haproxy does not include DTrace probes by any chance right? :)
No, and I have no idea how this works either. But if you feel like it
can provide some value and be done without too much effort, feel free
to try :-)
Here is a proof of
Hi John,
On Wed, May 07, 2014 at 06:14:13PM +0200, John-Paul Bader wrote:
Ok,
I have just built haproxy with your patches like this:
gmake TARGET=freebsd USE_PCRE=1 USE_OPENSSL=1 USE_ZLIB=1
When trying to start haproxy it failed with:
[ALERT] 126/160108 (25333) : Unable to allocate
Hi Vincent,
On Wed, May 07, 2014 at 06:35:06PM +0200, Vincent Bernat wrote:
??? 7 mai 2014 11:15 +0200, Willy Tarreau w...@1wt.eu :
haproxy does not include DTrace probes by any chance right? :)
No, and I have no idea how this works either. But if you feel like it
can provide some
My fix is broken, it should only show a warning and fallback on private
cache, i've just pointed the issue.
I will try to send you a workarounf patch soon.
Emeric
original message-
De: John-Paul Bader john-paul.ba...@wooga.net
A: Willy Tarreau w...@1wt.eu
Copie
Hmm yeah I noticed from what you wrote in the mail and by reading
through the patch - but still it confirmed that the shared pthread thing
was not available on FreeBSD right?
Would I also need to compile with USE_PRIVATE_CACHE=1 or would you patch
take care of that?
When it uses the private
❦ 7 mai 2014 22:19 +0200, Willy Tarreau w...@1wt.eu :
Here is a proof of concept. To test, use `make TARGET=linux2628
USE_DTRACE=1`. On Linux, you need systemtap-sdt-dev or something like
that. Then, there is a quick example in example/haproxy.stp.
Interesting, but just for my
❦ 7 mai 2014 22:56 +0200, Vincent Bernat ber...@luffy.cx :
So the main interest of those probes are:
* low overhead, they can be left in production to be here when you
really need them
And you enable/disable them while the program is running.
--
panic (No CPUs found. System
On Wed, May 07, 2014 at 10:59:43PM +0200, Vincent Bernat wrote:
??? 7 mai 2014 22:56 +0200, Vincent Bernat ber...@luffy.cx :
So the main interest of those probes are:
* low overhead, they can be left in production to be here when you
really need them
And you enable/disable them
Hi John,
On Wed, May 07, 2014 at 10:54:33PM +0200, John-Paul Bader wrote:
Hmm yeah I noticed from what you wrote in the mail and by reading
through the patch - but still it confirmed that the shared pthread thing
was not available on FreeBSD right?
Yes that's it. Old freebsd code did not
Hey,
I'm currently attempting to replace our commercial Loadbalancer with SSL
termination with haproxy. I'm running it on FreeBSD 9.2 Stable.
We have thousands of requests per second and for a while everything runs
extremely smooth. No queues are running full, machine load is at 0.5,
Hi,
I'm currently attempting to replace our commercial Loadbalancer with SSL
termination with haproxy. I'm running it on FreeBSD 9.2 Stable.
We have thousands of requests per second and for a while everything runs
extremely smooth. No queues are running full, machine load is at 0.5,
Hi,
On Tue, May 06, 2014 at 12:02:59PM +0200, Lukas Tribus wrote:
Hi,
I'm currently attempting to replace our commercial Loadbalancer with SSL
termination with haproxy. I'm running it on FreeBSD 9.2 Stable.
We have thousands of requests per second and for a while everything runs
Hey,
I will do more elaborate test runs in the next couple of days. I will
create traces with ktrace which is not as nice as strace but at least
will provide more context. Is there anything in particular you'd be
interested in like only syscalls?
Meanwhile I have build haproxy with debug
Hi John-Paul,
On Tue, May 06, 2014 at 11:57:08PM +0200, John-Paul Bader wrote:
Hey,
I will do more elaborate test runs in the next couple of days.
No problem.
I will
create traces with ktrace which is not as nice as strace but at least
will provide more context. Is there anything in
On Mon, Jul 22, 2013 at 06:07:03PM +0200, Mark Janssen wrote:
The setup has been running for a few days now (still with nbproc 1) and is
performing admirably.
I'd say this issue is fixed/resolved.
Excellent, thank you very much for your feedback Mark. I'm merging the
patches then.
Best
The setup has been running for a few days now (still with nbproc 1) and is
performing admirably.
I'd say this issue is fixed/resolved.
Thanks!
On Fri, Jul 19, 2013 at 9:58 PM, Mark Janssen maniac...@gmail.com wrote:
I've applied the patches and am running with the new version now. I'll let
I've applied the patches and am running with the new version now. I'll let
it run overnight (with nbproc back at 1...).
I'll probably switch back to nbproc 1 in the morning, as traffic starts
ramping up.
Mark
On Thu, Jul 18, 2013 at 10:42 PM, Willy Tarreau w...@1wt.eu wrote:
Hi Mark,
OK I
On Fri, Jul 19, 2013 at 09:58:19PM +0200, Mark Janssen wrote:
I've applied the patches and am running with the new version now. I'll let
it run overnight (with nbproc back at 1...).
I'll probably switch back to nbproc 1 in the morning, as traffic starts
ramping up.
Thank you Mark, then I'm
Hi Mark,
OK I could reproduce, debug and fix. It was a tough one, really...
More a problem of internal semantics than anything else, so I had
to test several possibilities and study their impacts and the corner
cases. In the end we get something that's fixed and better :-)
The issue was mostly
Hi Willy,
This explains why this only happens for short durations (at most the duration
of a client timeout).
Good to hear you pinpointed this.
What is important is to know that CPU usage aside, there is no loss of
information nor service. Connections are correctly handled
Mark early
On Wed, Jul 17, 2013 at 08:16:18PM +0200, Lukas Tribus wrote:
Hi Willy,
This explains why this only happens for short durations (at most the
duration
of a client timeout).
Good to hear you pinpointed this.
What is important is to know that CPU usage aside, there is no loss
?
I haven't tried yet... and can't really test this currently
Do you have the possibility to run without nbproc?
We were running without nbproc, and saw the 100% cpu load (though this was
still on dev18). We switched to nbproc=7 to work-around the 100% cpu load,
as I was told
I've tried with 'noepoll' and 'nosplice' ... and this seems to have solved
the cpu spikes... though the base-load is now a lot higher, due to using
poll instead of epoll.
Next I tried with noepoll, but with splice enabled. This resulted in the
same higher base-load, but still the occasional peak
On Fri, Jul 12, 2013 at 5:30 PM, Tomas Pospisek t...@sourcepole.ch wrote:
Am 11.07.2013 11:45, schrieb Mark Janssen:
I did see large amounts of sequential epoll_wait calls in the processes
with 100% cpu load, and not with the other processes.
epoll_wait(0, {}, 200, 0) = 0
the possibility to run without nbproc?
We were running without nbproc, and saw the 100% cpu load (though this was
still on dev18). We switched to nbproc=7 to work-around the 100% cpu load,
as I was told that the sites weren't responsive at the 100% cpu load times
(though I couldn't reproduce this)
How
Am 11.07.2013 11:45, schrieb Mark Janssen:
I've noticed that the HAProxy processes occasionally jump to 100% cpu
load, while the load before and after these peaks is only 3-5%, and the
traffic is also the same as outside of these cpu-peaks.
I saw a thread about this earlier (april/may
Hi Mark,
Hi list...
I've noticed that the HAProxy processes occasionally jump to 100% cpu
load, while the load before and after these peaks is only 3-5%, and the
traffic is also the same as outside of these cpu-peaks.
I saw a thread about this earlier (april/may), which concluded
Hi list...
I've noticed that the HAProxy processes occasionally jump to 100% cpu load,
while the load before and after these peaks is only 3-5%, and the traffic
is also the same as outside of these cpu-peaks.
I saw a thread about this earlier (april/may), which concluded that there
was a bug
56 matches
Mail list logo