New 502 Bad Gateway
-
With respect to the workaround that involves editing /usr/local/www/pfblockerng/www/index.php, I am not a PHP programmer, so I may be speaking out of turn, but I noticed that in the relevant code section two locks are acquired and never released:
if (!empty($pfb_query)) { // Increment DNSBL Alias Counter $dnsbl_info = '/var/db/pfblockerng/dnsbl_info'; if (($handle = @fopen("{$dnsbl_info}", 'r')) !== FALSE) { flock($handle, LOCK_EX); $pfb_output = @fopen("{$dnsbl_info}.bk", 'w'); flock($pfb_output, LOCK_EX); // Find line with corresponding DNSBL Aliasname while (($line = @fgetcsv($handle)) !== FALSE) { if ($line[0] == $pfb_query) { $line[3] += 1; } @fputcsv($pfb_output, $line); } @fclose($pfb_output); @fclose($handle); @rename("{$dnsbl_info}.bk", "{$dnsbl_info}"); } }Referring to the PHP documentation for the flock function, apparently this used to be okay because the locks were implicitly released when the file handles were closed, but that is no longer so starting from version 5.3.2: https://secure.php.net/manual/en/function.flock.php
That said, I don't know if the update from 2.3.4 to 2.4.0 happened to jump from a pre-5.3.2 version of PHP to a post-5.3.2 version. But regardless, I assume that explicitly releasing locks couldn't hurt anything. Also, the return value of one of the two fopen calls is checked but the other is not. I'm going to try the following modification on my system, which is still UFS, and will report back:
if (!empty($pfb_query)) { // Increment DNSBL Alias Counter $dnsbl_info = '/var/db/pfblockerng/dnsbl_info'; if (($handle = @fopen("{$dnsbl_info}", 'r')) !== FALSE) { if(flock($handle, LOCK_EX)) { if (($pfb_output = @fopen("{$dnsbl_info}.bk", 'w')) !== FALSE) { if(flock($pfb_output, LOCK_EX)) { // Find line with corresponding DNSBL Aliasname while (($line = @fgetcsv($handle)) !== FALSE) { if ($line[0] == $pfb_query) { $line[3] += 1; } @fputcsv($pfb_output, $line); } flock($pfb_output, LOCK_UN); } @fclose($pfb_output); } flock($handle, LOCK_UN); } @fclose($handle); @rename("{$dnsbl_info}.bk", "{$dnsbl_info}"); } }And sorry for the lousy formatting; I tried pasting several times and couldn't manage to get the indentation non-wonky.
-
It looks like pfSense has been using PHP 5.3.2 since around version 2.1 (the change log for 2.1 https://doc.pfsense.org/index.php/2.1_New_Features_and_Changes just says "PHP to 5.3.x"). So technically as far back as then all locks acquired with PHP's flock should have been manually released. Nevertheless, maybe some other change in 2.4.0 resulted in the lack of these explicit releases causing trouble where they had not before. I have two pfSense machines running 2.4.0 with pfBlockerNG and my modification now so I'll just see how they fare.
-
It looks like pfSense has been using PHP 5.3.2 since around version 2.1 (the change log for 2.1 https://doc.pfsense.org/index.php/2.1_New_Features_and_Changes just says "PHP to 5.3.x"). So technically as far back as then all locks acquired with PHP's flock should have been manually released. Nevertheless, maybe some other change in 2.4.0 resulted in the lack of these explicit releases causing trouble where they had not before. I have two pfSense machines running 2.4.0 with pfBlockerNG and my modification now so I'll just see how they fare.
You are correct. I just ran 'php -v' on my 2.3.4 system:
PHP 5.6.31 (cli) (built: Jul 14 2017 19:45:37) Copyright (c) 1997-2016 The PHP Group Zend Engine v2.6.0, Copyright (c) 1998-2016 Zend Technologies with Zend OPcache v7.0.6-dev, Copyright (c) 1999-2016, by Zend Technologies with Xdebug v2.5.0, Copyright (c) 2002-2016, by Derick Rethans with Suhosin v0.9.38, Copyright (c) 2007-2015, by SektionEins GmbH -
BBcan177 is currently working on an update to move this to a sqlite database, and get it off the text file. So this should never be an issue in the future.
All the effected code does is keep the updates for the desktop widget. Omitting the code does not break pfBlockerNG it only affects the widget reporting.
-
I pushed some experimental changes today that are designed to help track down the issue of gateway timeouts occurring when the IPSec dashboard widget is displayed. Those changes will be in the next 2.4 snapshot (probably early Oct 18th)
I would appreciate any feedback Better/worse/no change . .
Thanks
-
I also get the 502 Bad Gateway error when trying to access the pfSense Web GUI. I can logon a SSH session. But it hangs after entering my password. I noticed it about 24 hours after upgrade to 2.4. I use pfBlockerNG. Thanks for the information in this thread. I will continue to monitor for updates to the issue.
-
I pushed some experimental changes today that are designed to help track down the issue of gateway timeouts occurring when the IPSec dashboard widget is displayed.
I, myself never had the IPSEC dashboard widget displayed when I was getting the 502 Bad Gateway issue. Was this problem somehow thought to be related? I don't use IPSEC at all.
-
If it is related to memory or a connection or network queue, then in particular the output of these could be helpful:
/usr/bin/netstat -Ln /usr/bin/netstat -xn /usr/sbin/swapinfo -h /usr/bin/top | /usr/bin/head -n7 /bin/ps uxawwd /usr/bin/sockstatAttach the output in a text file as it will be too large to put inline on a forum post.
I believe that I am experiencing this problem; I have attached the requested files to confirm. I had disabled pfBlockerNG and found the system locked up again this morning; however, I had not disabled DNSBL.
-
I believe that I am experiencing this problem; I have attached the requested files to confirm. I had disabled pfBlockerNG and found the system locked up again this morning; however, I had not disabled DNSBL.
The output you posted confirms you are hitting this issue, it's definitely the pfBlocker DNSBL process getting stuck like others here in the thread.
-
I pushed some experimental changes today that are designed to help track down the issue of gateway timeouts occurring when the IPSec dashboard widget is displayed.
I, myself never had the IPSEC dashboard widget displayed when I was getting the 502 Bad Gateway issue. Was this problem somehow thought to be related? I don't use IPSEC at all.
In the past, the IPsec widget has been a different source of 502 errors. So it's possible that someone looking at this thread might be hitting that issue and not the DNSBL issue.
-
I have not had any issues but based on this post I have disabled DNSBL and pfblocker until the next update. Bummer since this is one of my favorite parts of using pfsense.
-
Can anyone confirm that the issue is not present provided that DNSBL is disabled? Is that all that needs to be done until upgrade?
-
The file is at this location
/usr/local/www/pfblockerng/www/index.php
cd /usr/local/www/pfblockerng/www/cp index.php index.old = do this so you have a copy of the original before you remove the section.Edit index.php with your favorite editor, and remove this section at the bottom.
if (!empty($pfb_query)) { // Increment DNSBL Alias Counter $dnsbl_info = '/var/db/pfblockerng/dnsbl_info'; if (($handle = @fopen("{$dnsbl_info}", 'r')) !== FALSE) { flock($handle, LOCK_EX); $pfb_output = @fopen("{$dnsbl_info}.bk", 'w'); flock($pfb_output, LOCK_EX); // Find line with corresponding DNSBL Aliasname while (($line = @fgetcsv($handle)) !== FALSE) { if ($line[0] == $pfb_query) { $line[3] += 1; } @fputcsv($pfb_output, $line); } @fclose($pfb_output); @fclose($handle); @rename("{$dnsbl_info}.bk", "{$dnsbl_info}"); } }Just to add to this thread, I can confirm that the above fix worked for me. I had this issue after pushing out the upgrade to 2.4 and followed the post above (I commented the lines out rather than deleting them). Since then it has been stable and all pfSense routers in my environment have stopped giving the bad gateway error.
-
I am going to make a PR shortly to hopefully fix this issue….
You could download these two files from my Github gist to get the updated code:
fetch -o /usr/local/www/pfblockerng/www/index.php "https://gist.githubusercontent.com/BBcan177/9f9c8e62b166cee07ad16cd4ff59103c/raw" fetch -o /usr/local/pkg/pfblockerng/pfblockerng.inc "https://gist.githubusercontent.com/BBcan177/7ff15715be0f02afdbe0a00c676aedce/raw"You will need to restart the DNSBL Service for this to take effect.
A reboot is not required, but would be recommended.
-
If it is related to memory or a connection or network queue, then in particular the output of these could be helpful:
/usr/bin/netstat -Ln /usr/bin/netstat -xn /usr/sbin/swapinfo -h /usr/bin/top | /usr/bin/head -n7 /bin/ps uxawwd /usr/bin/sockstatAttach the output in a text file as it will be too large to put inline on a forum post.
@jimp From my topic yesterday https://forum.pfsense.org/index.php?topic=138283.0
You referred me to get these command outputs when it happened again and it finally has. I have attached the text file. Let me know if you need anything else and thanks again.
[pfSense Command Outputs.txt](/public/imported_attachments/1/pfSense Command Outputs.txt)
-
I am going to make a PR shortly to hopefully fix this issue….
You could download these two files from my Github gist to get the updated code:
fetch -o /usr/local/www/pfblockerng/www/index.php "https://gist.githubusercontent.com/BBcan177/9f9c8e62b166cee07ad16cd4ff59103c/raw" fetch -o /usr/local/pkg/pfblockerng/pfblockerng.inc "https://gist.githubusercontent.com/BBcan177/7ff15715be0f02afdbe0a00c676aedce/raw"You will need to restart the DNSBL Service for this to take effect.
A reboot is not required, but would be recommended.
Done and works fine so far as I can tell. Good work!
Side note-have been watching for V3. :-) -
You referred me to get these command outputs when it happened again and it finally has. I have attached the text file. Let me know if you need anything else and thanks again.
Looks like the same pfblocker dnsbl issue as others in the thread. There is a pfBlocker update that will be out momentarily that should help, if not fix it entirely.
-
I have not had any issues but based on this post I have disabled DNSBL and pfblocker until the next update. Bummer since this is one of my favorite parts of using pfsense.
I have been up and running for ~10 hours since disabling DNBL. Previsouly, I had just pfBlocker disabled - and it locked up overnight.
-
So I have disabled pfblocker, that fixed the crashing or hang that was occurring every 8 hours or so. However now the gui and ssh become non responsive after 8 hours or so.
I'm going to try fresh install with ZFS to see if the problems continue.
-
Very interesting. I did not have this 502 Bad Gateway Issue after upgrading to 2.4.0-RELEASE. (Uptime since Upgrade 1 Day 18 Hours).
I just installed now the new Update (Upgrading pfSense-pkg-pfBlockerNG from 2.1.1_11 to 2.1.2)My Setup:
Kettop Mi19N (32GB SSD UFS Filesystem + 8GB RAM)
PFBlockerNG + DNSBL enabled
MultiWAN (Failover)
IPSEC + OpenVPN
