https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Graham Leggett changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #74 from Tom Drake ---
Carsten, thanks for your reply.
Do you have any suggestions for me as to how to get around the issue I'm
experiencing?
Thanks.
--
You are receiving this mail because:
You are the assignee for the bug.
-
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #73 from Carsten Wolff ---
Hi Tom,
(In reply to Tom Drake from comment #72)
> I believe that I am encountering this same issue.
that is unlikely. You're using mod_proxy_ajp, which does not share much with
mod_proxy_connect. Also,
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #72 from Tom Drake ---
I believe that I am encountering this same issue.
I had reported my issue on server fault here:
https://serverfault.com/questions/949259/apache-buffers-tomcat-responses-and-closes-connection-to-tomcat-when-r
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #71 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #70)
> I tries to keep the best of our respective proposals (no buffering nor code
> duplication...), if it works we should find an agreement right? :)
Hi Yann,
I ap
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35783|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #69 from Yann Ylavic ---
(In reply to Yann Ylavic from comment #68)
> You don't seem to be using attachment 35620 [details] here (hence
> FlushMinThreshold).
This relates to comment 65, not Mohamed's patch.
--
You are receiving th
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #68 from Yann Ylavic ---
You don't seem to be using attachment 35620 here (hence FlushMinThreshold).
If this is the case, you at least need to:
#define THRESHOLD_MIN_WRITE 1
in "server/core_filters.c".
Otherwise, the core output
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Mohamed El Housseine changed:
What|Removed |Added
Attachment #35776|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #66 from Mohamed El Housseine ---
(In reply to Yann Ylavic from comment #60)
Hi Yann again,
thank you very much for for your detailed reply.
> The issue is that ap_proxy_transfer_between_connections() is a loop, and
> once it has c
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #65 from Carsten Wolff ---
Created attachment 35786
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35786&action=edit
log of a single process entering a busy loop
(In reply to Yann Ylavic from comment #63)
> (In reply to Yan
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #64 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #62)
> Hmm, I can't reproduce, can you please provide the error log when this
> happens?
Will do. In the meantime, another observation: Because your patch v12 is
cons
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #63 from Yann Ylavic ---
(In reply to Yann Ylavic from comment #62)
> (In reply to Carsten Wolff from comment #61)
> >
> > unfortunately not, it creates busy loops again, when I:
> > - start 100 transfers "simultaniously"
> > - qui
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #62 from Yann Ylavic ---
(In reply to Carsten Wolff from comment #61)
>
> unfortunately not, it creates busy loops again, when I:
> - start 100 transfers "simultaniously"
> - quickly interrupt the rsync clients (SIGINT)
Hmm, I can
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #61 from Carsten Wolff ---
Hi Yann,
(In reply to Yann Ylavic from comment #56)
> Created attachment 35783 [details]
> Handle POLLOUT in proxy_connect (v12)
>
> Does it work for you?
unfortunately not, it creates busy loops again,
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #60 from Yann Ylavic ---
Hi Mohamed,
(In reply to Mohamed El Housseine from comment #58)
>
> In its second call (for handling remaining data in input filters) it can
> return:
>
> APR_SUCCESS
> APR_STATUS_IS_INCOMPLETE
> ERROR
>
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #59 from Mohamed El Housseine ---
(In reply to Mohamed El Housseine from comment #58)
sorry
> For ensuring this, extending the if-condition for handling
> data_in_output_filters with "&& !APR_STATUS_IS_EAGAIN(rv)" must be enought.
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #58 from Mohamed El Housseine ---
Hi Yann,
thank you for your feedback!
> It also keeps the ap_proxy_transfer_between_connections() logic in one
> place, to avoid duplicating code (and error handling, which seems to be
> missing in
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35784|Diff from v11 to v12|Incremental patch from v11
descri
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #57 from Yann Ylavic ---
Created attachment 35784
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35784&action=edit
Diff from v11 to v12
--
You are receiving this mail because:
You are the assignee for the bug.
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35738|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Mohamed El Housseine changed:
What|Removed |Added
Attachment #35774|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #54 from Mohamed El Housseine ---
Created attachment 35774
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35774&action=edit
patch for fixing the problem of remaining data in input ctx
Hi yan,
Hi Carsten,
i could fix that p
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #53 from Carsten Wolff ---
BTW,
(In reply to Mohamed El Housseine from comment #48)
> #b: socket_bucket_read allocates a new bucket and _overwrites_ the requested
> len with APR_BUCKET_BUFF_SIZE, before passing it to apr_socket_rec
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #52 from Carsten Wolff ---
Created attachment 35772
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35772&action=edit
try draining input buffers
this patch goes on top of
https://bz.apache.org/bugzilla/attachment.cgi?id=3573
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Carsten Wolff changed:
What|Removed |Added
CC||elh.moha...@gmail.com
--
You are rece
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #51 from Yann Ylavic ---
Yes sorry, I made a shortcut. POLLOUT + a "drain" flag to indicate that the
POLLIN code should be entered in this case too...
--
You are receiving this mail because:
You are the assignee for the bug.
-
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #50 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #49)
> The idea is to "drain" the input chain by not clearing POLLOUT (on the other
> side) until done.
How would that work? After POLLOUT we only call ap_core_output
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #49 from Yann Ylavic ---
Thanks Mohamed for the nice analysis, yes the issue is that there may still be
data in the input filter chain (which we didn't read yet) when the output
filter chain becomes full and we abort the transfer (A
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #48 from Mohamed El Housseine ---
After driving many tests with the data provided by Carsten, I was able to
figure out, where the unsent-but-read data sits during the stall.
For simplification I will reduce the proxy flow to the fo
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #47 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #46)
> The pcap shows rsynd sending data at the (timeout/premature) end of
> connection which is an error message (on the socket???).
> Something's fishy there...
Not
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #46 from Yann Ylavic ---
Created attachment 35739
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35739&action=edit
rsync timeout: pcap and httpd log (with v11)
I used three different (linux) boxes for rsync, the proxy and r
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35623|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #44 from Carsten Wolff ---
Created attachment 35737
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35737&action=edit
fixes a busy loop trying to shutdown a disconnected socket
(In reply to Carsten Wolff from comment #43)
>
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #43 from Carsten Wolff ---
(In reply to Carsten Wolff from comment #42)
> I have a feeling I should repeat Tests 1 and 2 without SSL to see if SSL is
> necessary for all the stalls to be reproduced. I'll let you know.
So I did:
Te
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #42 from Carsten Wolff ---
Created attachment 35639
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35639&action=edit
pcaps of a stall in rsync handshake (Test4)
Hello Yann, I wish you a happy new year!
I did some more test
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35621|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35513|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35431|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #38 from Carsten Wolff ---
(In reply to Carsten Wolff from comment #37)
> (In reply to Yann Ylavic from comment #36)
> > If this is the case, I'd suggest using attachment 35513 [details] only for
> > now, that's less performances (t
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #37 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #36)
> Sorry for the delay Carsten, just to clarify, you are running httpd with
> attachment 35513 [details] [Handle POLLOUT in proxy_connect (v8)] and
> attachment 35
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #36 from Yann Ylavic ---
Sorry for the delay Carsten, just to clarify, you are running httpd with
attachment 35513 [Handle POLLOUT in proxy_connect (v8)] and attachment 35447 [
Read buffer size and flush threshold tuning], no other
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #35 from Carsten Wolff ---
Hum. It seems to be even worse. Now that proxy_connect works for my case with
these patches, everything else seems to be broken, even the normal case of GET
for static files produces incomplete files.
To t
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Carsten Wolff changed:
What|Removed |Added
CC||carsten.wo...@dsa.de
--- Comment #34 f
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #33 from Carsten Wolff ---
Tests with v8 look good with SSL. Thanks! :-)
--
You are receiving this mail because:
You are the assignee for the bug.
-
To unsubscrib
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35512|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35446|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #30 from Carsten Wolff ---
(In reply to Carsten Wolff from comment #29)
> Hi, unfortunately, using patch#35446 and patch#35447, there is another
> problem, when I change the scenario by making the client connection through
> SSL. In
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #29 from Carsten Wolff ---
Hi, unfortunately, using patch#35446 and patch#35447, there is another problem,
when I change the scenario by making the client connection through SSL. In that
case, the transfer stalls very quickly within
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #28 from Carsten Wolff ---
Thanks Yann for your patience and all your work! I have some last observations
to share:
- When combining https://bz.apache.org/bugzilla/attachment.cgi?id=35427 and
https://bz.apache.org/bugzilla/attachme
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #27 from Yann Ylavic ---
The logs from comment 21 and comment 26 look normal, no timeout triggered,
proxy_connect_handler() seems to be waiting for sockets to be ready to do its
job.
(In reply to Carsten Wolff from comment #26)
> I
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #26 from Carsten Wolff ---
I added this just at the end of the poll() loop:
ap_log_rerror(APLOG_MARK, APLOG_TRACE2, 0, r,
"@END: "
"[%s: POLLIN: %d POLLOUT: %d POLLHUP: %d data_in_filters: %d] "
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #25 from Yann Ylavic ---
(In reply to Yann Ylavic from comment #24)
> Created attachment 35447 [details]
> Read buffer size and flush threshold tuning
Also not note that send_brigade_blocking() is now a bit misnamed, it's rather
se
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35427|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #23 from Yann Ylavic ---
(In reply to Carsten Wolff from comment #21)
>
> It gives me a hard time to reproduce the original issue. :-) But, it now
> blocks at the end of the transfer in mod_proxy_connect's poll() for 300
> seconds
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #22 from Yann Ylavic ---
(In reply to Carsten Wolff from comment #21)
>
> Maybe there should be a APR_NONBLOCK_WRITE flag one could
> pass to ap_pass_brigade() (and further down to send_brigade_blocking) to
> absolutely make sure,
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #21 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #19)
> Thanks for the patch Carsten, but there can't be an exception for
> send_brigade_blocking() to block when the conditions are met, the core
> output filter must
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35432|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #19 from Yann Ylavic ---
Thanks for the patch Carsten, but there can't be an exception for
send_brigade_blocking() to block when the conditions are met, the core output
filter must control the memory buckets in flight (set aside).
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Carsten Wolff changed:
What|Removed |Added
Attachment #35442|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #17 from Carsten Wolff ---
Created attachment 35442
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35442&action=edit
suppress poll with long timeout in filter code
Hi again,
this patch, ontop of https://bz.apache.org/bugzi
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #16 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #15)
> Created attachment 35432 [details]
> Handle POLLOUT in proxy_connect (v5)
>
> Please use v5 instead, v4 can't work before it does not play well with
> EAGAIN.
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35430|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #14 from Carsten Wolff ---
Created attachment 35431
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35431&action=edit
shortening the poll timeout for send_brigade_blocking
I tried this patch on top of your patches, but unfor
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35426|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #13 from Yann Ylavic ---
Created attachment 35430
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35430&action=edit
Handle POLLOUT in proxy_connect (v4)
Thanks for the traces, I missed that proxy_connect was FLUSHing data.
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #12 from Carsten Wolff ---
Created attachment 35429
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35429&action=edit
trace6 error log of a single test
(In reply to Yann Ylavic from comment #11)
> Could you run the test with
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #11 from Yann Ylavic ---
Could you run the test with LogLevel trace6 and provide the error_log?
--
You are receiving this mail because:
You are the assignee for the bug.
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #10 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #9)
> Could you please test it and see if it helps for this case too?
> It should be applied in addition to attachment 35426 [details] (v3) because
> proxy_connect nee
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #9 from Yann Ylavic ---
Created attachment 35427
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35427&action=edit
File read buffer and flush threshold tuning
Thanks for testing Carsten.
I think I see what happens now, the
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #8 from Carsten Wolff ---
(In reply to Yann Ylavic from comment #7)
> We have to stop reading on the other side when writing blocks on one side.
> How does this new version work?
Not better, I'm afraid:
Thread 2 (Thread 0x7f534bff
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35425|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #6 from Carsten Wolff ---
First of all, thanks for the very quick turnaround!
Unfortunately, the patch doesn't fix the problem. Actually, I have played with
similar approaches for changing proxy_connect_handler() before, like doing
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic changed:
What|Removed |Added
Attachment #35424|0 |1
is obsolete|
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #4 from Yann Ylavic ---
Created attachment 35424
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35424&action=edit
Handle POLLOUT in proxy_connect
Could you please try this patch on latest 2.4.x?
--
You are receiving this
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #3 from cars...@wolffcarsten.de ---
One more remark: I selected version 2.4.25 in the bug header, because I used
that (debian) version for the reproducer. Effected versions are actually all
versions, including 2.2 branch, 2.4 branch
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #2 from cars...@wolffcarsten.de ---
Created attachment 35423
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35423&action=edit
script for reproducing the bug
--
You are receiving this mail because:
You are the assignee for t
https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
--- Comment #1 from cars...@wolffcarsten.de ---
Created attachment 35422
--> https://bz.apache.org/bugzilla/attachment.cgi?id=35422&action=edit
rsync daemon config
--
You are receiving this mail because:
You are the assignee for the bug.
---
78 matches
Mail list logo