| Author |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
 Posted: Fri 24 Feb '12 4:46 Post subject: Difference mod_cgi versions |
 |
|
| Should it upgrade from v2.3.6 to v2.3.7 now? Is there any error running at win32 servers? |
|
| Back to top |
|
James Blond
Moderator
Joined: 19 Jan 2006
Posts: 7371
Location: Germany, Next to Hamburg
|
| Posted: Fri 24 Feb '12 11:46 Post subject: |
|
|
If you can't live with the issues the plain 2.3.6 has than maybe. I still use 2.3.6d from ApacheHaus which includes that patches but not the "new" stuff.
I haven't tested 2.3.7 under load yet. |
|
| Back to top |
|
admin
Site Admin
Joined: 15 Oct 2005
Posts: 692
|
| Posted: Fri 24 Feb '12 11:49 Post subject: |
|
|
Up to you to upgrade from the patch 2.3.6 here or AH, see the Changes.
No issues seen, reported on win32.
Steffen
EDIT: oops, James posted at the same time  |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Sat 25 Feb '12 2:44 Post subject: |
|
|
I saw the report https://issues.apache.org/bugzilla/show_bug.cgi?id=52695 .
His operation is Linux.
But,no one reports that under win32 environment.
My operation is win32 2003 server.
Does the mod_fcgid 2.3.7 fix the v2.3.6 bugs which ever had? |
|
| Back to top |
|
glsmith
Moderator
Joined: 16 Oct 2007
Posts: 2268
Location: Sun Diego, USA
|
| Posted: Sat 25 Feb '12 8:16 Post subject: |
|
|
Yes it does, PR 50309 and some others
see change log |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Tue 28 Feb '12 11:48 Post subject: |
|
|
thanks.
Does the error get fix below either?
| Code: |
Changes with mod_fcgid 2.3.6
*) SECURITY: CVE-2010-3872 (cve.mitre.org)
Fix possible stack buffer overwrite. Diagnosed by the reporter.
PR 49406. [Edgar Frank <ef-lists email.de>]
*) Change the default for FcgidMaxRequestLen from 1GB to 128K.
Administrators should change this to an appropriate value based on
site requirements. [Jeff Trawick]
*) Allow FastCGI apps more time to exit at shutdown before being
forcefully killed. [Jeff Trawick]
*) Correct a problem that resulted in FcgidMaxProcesses being ignored
in some situations. PR 48981. [<rkosolapov gmail.com>]
*) Fix the search for processes with the proper vhost config when
ServerName isn't set in every vhost or a module updates
r->server->server_hostname dynamically (e.g., mod_vhost_cdb)
or a module updates r->server dynamically (e.g., mod_vhost_ldap).
[Jeff Trawick]
*) FcgidPassHeader now maps header names to environment variable names
in the usual manner: The header name is converted to upper case and
is prefixed with HTTP_. An additional environment variable is
created with the legacy name. PR 48964. [Jeff Trawick]
*) Allow processes to be reused within multiple phases of a request
by releasing them into the free list as soon as possible.
[Chris Darroch]
*) Fix lookup of process command lines when using FcgidWrapper or
access control directives, including within .htaccess files.
[Chris Darroch]
*) Resolve a regression in 2.3.5 with httpd 2.0.x on some Unix platforms;
ownership of mutex files was incorrect, resulting in a startup failure.
PR 48651. [Jeff Trawick, <pservit gmail.com>]
*) Return 500 instead of segfaulting when the application returns no output.
[Tatsuki Sugiura <sugi nemui.org>, Jeff Trawick]
*) In FCGI_AUTHORIZER role, avoid spawning a new process for every
different HTTP request. [Chris Darroch]
|
| glsmith wrote: | Yes it does, PR 50309 and some others
see change log |
|
|
| Back to top |
|
James Blond
Moderator
Joined: 19 Jan 2006
Posts: 7371
Location: Germany, Next to Hamburg
|
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Tue 28 Feb '12 13:12 Post subject: |
|
|
Not that.
I mean,Does this 2.3.7 release from apachelounge can replace the mod_fcgid2.3.6d of apachehaus that has ever fixed issues?
|
|
| Back to top |
|
James Blond
Moderator
Joined: 19 Jan 2006
Posts: 7371
Location: Germany, Next to Hamburg
|
| Posted: Tue 28 Feb '12 21:42 Post subject: |
|
|
| Yepp the 2.3.7 includes all that patches and some stuff I haven't reviewed yet. |
|
| Back to top |
|
glsmith
Moderator
Joined: 16 Oct 2007
Posts: 2268
Location: Sun Diego, USA
|
| Posted: Tue 28 Feb '12 23:07 Post subject: |
|
|
Comparison of issues fixed in the various versions.
| Code: |
2.3.6d 2.3.7
[1] PR 50309 PR 50309
[2] r1037483 r1037483
[3] r1037727 r1037727
[4] PR 51020 --------
[5] PR 51560 is same as 51020
[6] PR 51749
[7] PR 51078
1. Graceful restart problem on Windows, our biggest
problem with 2.3.6
2. account for first process in class in the spawn
score
3. Fix regression in 2.3.6 which broke process
controls when using vhost-specific configuration.
4. Apache/mod_fcgid.so does not start in complex
Active Directory forest. ** Not in 2.3.7 **
5. Solve latency/cogestion of resolving effective
user file access rights when no such info is
desired, for config related filename stats.
6. Periodically clean out the brigades which are
pulling in the request body for handoff to the
fcgid child.
7. *** Introduce FcgidWin32PreventOrphans directive on
Windows to use OS Job Control Objects to
terminate all running fcgi's when the worker
process has been abruptly terminated.
** The only thing in 2.3.6d that is not in 2.3.7
looks to be PR 51020 dealing with mod_fcgid on
box with complex AD forest.
*** When I read this PR, in my view, this was
another way of dealing with the problem that
PR50309 fixes. The "Graceful Restart Problem"
wasn't just on graceful restarts, but also on
shutdown, it just got it's name because it was
most often seen during a graceful restart. Due
to this thinking I chose not to include it in
2.3.6d.
|
EDIT: *** It seems even though this PR is mentioned in the CHANGES-FCGID file, it is not there. Looking at the patch supplied with PR51078, I cannot find the code, nor the commit to add this code in svn. I want to have it removed from the changes file, but I do not want this patch added since I know PR50309 takes care of the root cause of this bug. Hmmmm, what to do?
Last edited by glsmith on Fri 02 Mar '12 21:35; edited 4 times in total |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Wed 29 Feb '12 11:39 Post subject: |
|
|
Get it.
thanks. |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Wed 29 Feb '12 12:19 Post subject: |
|
|
After upgrade to mod_fcgid2.3.7el,I change to the old version of mod_fcgid2.3.6d from apachehaus.
I find the v2.3.7,it will slow the web page connection time that v2.3.6d doesn't occur before.The internet explorer will appear error layout due to the slow connection time.
I don't know why.Maybe the new version has minor bug still.Or just like the user says,it can't undergo the more heavy loading?
To compare with the v2.3.6d and v2.3.7,the v2.3.6d provides the more fast connection time to explorer web page under the same server setting environment.
I use three internet explorers to test it.(IE9,firefox10.0.2,google chrome17)
The three kinds of explorers all lead to the same result that I mentioned above. |
|
| Back to top |
|
Steffen
Moderator
Joined: 15 Oct 2005
Posts: 3092
Location: Hilversum, NL, EU
|
| Posted: Wed 29 Feb '12 12:40 Post subject: |
|
|
Have no reports from other heavy traffic windows sites.
You say: Connection time longer. Does it happen all the time or when a new instance of the php-cgi.exe has to load ?
And what about mod_fcgid-2.3.6-P2 from here ?
Steffen |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Wed 29 Feb '12 14:10 Post subject: |
|
|
I think it happens wehn a new instance of php-cgi.exe.If you try the second time under the web cache,the page load soon than before.The explorer cache fasten the loading speed.
| Steffen wrote: | Have no reports from other heavy traffic windows sites.
You say: Connection time longer. Does it happen all the time or when a new instance of the php-cgi.exe has to load ? |
I don't test mod_fcgid2.3.6-p2 as yet.
I will try that if possible.
| Quote: |
And what about mod_fcgid-2.3.6-P2 from here ?
|
Is there any possibility to compile a new mod_fcgid2.3.7 that bases on the mod_fcgid2.3.6d?
Last edited by maskego on Thu 01 Mar '12 0:58; edited 1 time in total |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Wed 29 Feb '12 14:32 Post subject: |
|
|
I make a brief time to test mod_fcgid2.3.6p.
If use mod_fcgid2.3.6p,the connection time is slow as mod_fcgid2.3.7el shows.I get the same result under mod_fcgid2.3.6p and mod_fcgid2.3.7el approximately.It shows connection time out or can't show the web page sometimes.
In contrast,I change to mod_fcgid2.3.6d,the connection time becomes fast and web loading correct.
The test is under the same server setting environment via same internet explorers. |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Thu 01 Mar '12 9:25 Post subject: |
|
|
| Can apachehaus compile mod_fcgid2.3.7 that bases on mod_fcgid2.3.6d to download? |
|
| Back to top |
|
Steffen
Moderator
Joined: 15 Oct 2005
Posts: 3092
Location: Hilversum, NL, EU
|
| Posted: Thu 01 Mar '12 11:44 Post subject: |
|
|
Little strange your experience, no other reports. Only diff to P and d is the restart fix.
Have others the issue as maskego ?
Steffen |
|
| Back to top |
|
maskego
Joined: 16 Apr 2010
Posts: 238
|
| Posted: Thu 01 Mar '12 12:31 Post subject: |
|
|
steffen:
Thanks.
I feel strange too.
There is no error messages shows when use mod_fcgid2.3.7el.It just slows the web page connection and sendback time.I install apache2.2.22 that downloads from apachelounge.
Sometimes,it shows no page or page is shortly unavailable.
But,there is no such case after change back to mod_fcgid2.3.6d. |
|
| Back to top |
|
glsmith
Moderator
Joined: 16 Oct 2007
Posts: 2268
Location: Sun Diego, USA
|
| Posted: Thu 01 Mar '12 13:16 Post subject: |
|
|
| maskego wrote: | | Can apachehaus compile mod_fcgid2.3.7 that bases on mod_fcgid2.3.6d to download? |
You mean add in #5 & 6 from the list above? Have you thought it may be possible one of those 2 are what is causing your slow down?
Note: I did some research and have since edited that post. |
|
| Back to top |
|
James Blond
Moderator
Joined: 19 Jan 2006
Posts: 7371
Location: Germany, Next to Hamburg
|
| Posted: Thu 01 Mar '12 13:52 Post subject: |
|
|
@Gregg, yes you are right the patch from PR51078 has not been applied. Bill wanted to do that. But he hasn't done so.
PR51020 has been applied, but without the comments. It is now line 799 |
|
| Back to top |
|
