Apache graceful restarts and FastCGI - mod_fcgid: can't lock process table in pid

[rackaid]

This issue is becoming critical for many cPanel customers. Can cPanel recommend a workaround until they have a solution. This has been reported as fixed in upstream versions of Apache.

I have seen the following recommendations:

1. Switched to piped logs
2. Hard restart apache via cron

However, this does not address the problem of orphaned PHP processes already in existence.

Suggestion:
In our testing, we find on Apache 2.0 w/prefork MPM, the problem only exists with graceful restarts. Full restarts do not cause this.

Since cPanel supplies the restart scripts for Apache, can you supply one that will do a forceful restart always?

****UPDATE****
We just rebuilt Apache with the same profile as 2 days ago. But this time we cleared out the easyapache src tree to force a new download. This rebuild does not appear to suffer from the FastCGI issues.

With the prior build a graceful restart would not terminate fast-cgi processes. Now it does.


We need to let the system run for a bit more but this may have resolved it.

If we switch out the prior mod_fastcgi.so with the current one, the problem returns.

So you may want to try clearing easyapache src directory prior to any attempts to fix this issue.


****UPDATE 2****
Appears cPanel may have pushed a new EA build during our work. Try to rebuild your profile but make sure you have the latest EA by clearing out the src directory.

[ikillbill]

just applied EasyApache 3.11.2 , apache restart still does not kill orhpan PHP process.....
when this will be formally fixed please?

[cPanelTristan]

Hello,

Please reply to any open ticket or open up a ticket if any patch appears not to fix the issue. The only way to ensure development is aware if a specific patch doesn't function fully for resolving an issue would be to ensure the bug report ticket is the one used for reporting purposes.

Thanks!

[ikillbill]

bug report

#2464286

still NOT fixed

thanks

[jocthbr]

I'm freaking out with this issue!
Won't you fix it?

[GregoryBE`]

Any update concerning this issue yet?

[cPanelTristan]

Please try to update using the latest EasyApache to see the results. There were fixes placed into the most recent builds and some users have indicated that the issue appears resolved. If it is not resolved with the latest EasyApache, please submit a ticket because there is no way to investigate why the issue would still be occurring in your environment via the forums.

[Quemeros]

Old server got the problem... new server, with a fresh install, newest EasyApache, same result's... i don't stand where is the problem, but i will go to open another ticket.

[complexi]

@Quemeros

Just an FYI:

This suggestion by brainyron currently works for us:

*/5 * * * * pkill -9 -P 1 php

Not ideal as we don't want to depend on cron jobs to clean up orphaned php process.

Like your last post read, we'd rather have cPanel fix the bug.

complexi

[Quemeros]

@Complexi Thanks! I was use that exact command, is the best for the moment.

cPanel parcialy fix the problem if you have Log Rotation Log's active. I confirmed, but you CAN'T restart apache, because the process generated.

[Avanti]

We have used the below suggestion as a workaround and can confirm that our servers have been stable since this work around and speed seems comparable to FCGI to us.

We are so happy with this set-up that this is our new standard set-up. Configuration is also a lot simpler than FCGI. We didn't need to add any configuration settings into the Apache configuration editor.

For us this has been an all-round winner.

The best setup for fastest processing and lower memory usage would be to use DSO with mod_ruid2 rather than mod_fcgid for PHP. mod_ruid2 will not only run processes for PHP as the user, it will run Apache processes as the user for html and other pages. Everything under Apache that runs on the account will run as the user. DSO doesn't have the issues and configuration requirements that FCGI does.

[ikillbill]

just applied latest easyapache again, still not working
new ticket is issued again
#2511323

[mofolo]

Hi,

Process kill script is not a perfect solution.

I recomend to patch and replace current mod_fcgid with new one.

Bug: https://issues.apache.org/bugzilla/s...g.cgi?id=50309

Fix : https://issues.apache.org/bugzilla/a...82&action=diff

Feel free to download and change source code of fcgi module. After - install it to your apache ( replace existing ).
You will need to reinstall it each time after apache rebuild via EasyApache.

My patched module is attached ( mod_fcgid-2.3.6 )
For me this bug is fully fixed. Attachment 9341



Good luck ! Have a nice day ))

Confirmed to be a working fix.

This is what you have to do:

1. Download mod_fcgid from the link above.

2. Extract the .tar and use something like SSH/FTP to transfer the directory in your Apache root directory
For me running CentOS 5 it was /etc/httpd/

3. Use SSH and direct it to the folder you have just uploaded
For me it was.

# cd /etc/httpd/mod_fcgid-2.3.6/

4. Run the configuration file

# ./configure.apxs

5. Make and install

# make

# make install

6. Restart Apache

# /etc/init.d/httpd restart

This problem hasn't bothered me since (it's been 24 hours and i was having this problem every 2-3 hours).

I hope this helps someone else. I was pulling out my hair cause of this.

[Juanzo]

Should we use mofolo's solution or just add */5 * * * * pkill -9 -P 1 php as a cron, until cPanel confirms if next EA will fix this?

[cPanelTristan]

You can try mofolo's solution, but any module you build using apxs that adds to /usr/local/apache/modules/ folder needs to be backed up to /root folder and then re-added at the end of any EasyApache you run. If you don't have a /usr/local/cpanel/scripts/posteasyapache script where you copy the module back into /usr/local/apache/modules/ folder, your EasyApache recompile will load the other fcgi module that we'll replace it with.