New 502 Bad Gateway

dstroot

I think the "Can NOT be reverted cleanly" through me off. I just checked and mine and got the results you show so I applied the patch and it applied successfully. I will report if the 502s go away. Cheers!

fortrash

Same error here. It is quite annoying as my pfsense installation ran smoothly. Since 2.4 I have to schedule reboots.
Any workaround except commenting out the code?
Are the developer working on it? Otherwise I do test the competitor product.

Thanks,

PiBa

Fortrash, please try the patch: https://forum.pfsense.org/index.php?topic=137103.msg767259#msg767259

fortrash

Thanks any chance to help the developer?

PiBa

@fortrash:

Thanks any chance to help the developer?

How do you mean?

Patches are made by developers.. Testing if it indeed fixes the issue would help, but thats something 'you' (users that experience the actual issue) have to do..

fortrash

Thanks, I am testing the patch.

chrcoluk

guys check my post here

https://forum.pfsense.org/index.php?topic=110515.msg766964#msg766964

PatPend

Running pfsense 2.4.1-RELEASE with pfBlockerNG on 2.1.2_1 installed and also running OpenVPN server. When the Bad Gateway error happened, OpenVPN clients couldn't connect as well. Rebooting fixed it for now. Any news on when the patch might be rolled into an update? Thanks, Russ

PiBa

PatPend, any news on if the patch helps? ::) may i presume your running with pfBlocker dns blocklists enabled as well?

If there is no positive feedback on its results then there is no need to commit it right? I hope there are no new problems reported by the people that applied the patch, and that they can confirm the problem did not return or at the very least took longer to re-appear.. Would be nice that in a week time they could say it has been running stable..

In the mean time please feel free to apply the patch to your own installation and test it out as well.

PatPend

@PiBa:

PatPend, any news on if the patch helps? ::) may i presume your running with pfBlocker dns blocklists enabled as well?

If there is no positive feedback on its results then there is no need to commit it right? I hope there are no new problems reported by the people that applied the patch, and that they can confirm the problem did not return or at the very least took longer to re-appear.. Would be nice that in a week time they could say it has been running stable..

In the mean time please feel free to apply the patch to your own installation and test it out as well.

I applied the patch today, so far so good. I'm running with easylist and easylist privacy plus 28 custom entries to block smart TV traffic. Last time it took about 10 days before hitting the 502 (of course when it did I was out of town when it also took out OpenVPN ::) ).

BBcan177

@PiBa:

For those willing to give some new code a try i have made a few changes to the 'file locking' code of pfBlockerNG.. :)

Thanks PiBa, hopefully we get some feedback to see if this resolves this issue… and we can get it merged!

fortrash

Hello,

I had to restart my pfsense box again. The patch has already been installed. Should I enable any logging?
Let me know how and what I can provide to resolve the problem.

Regards,

PiBa

Hmm ok thanks..

Could you create a file /root/testpfb.sh with following content?
The 'on-mobi.com' is one of the websites mentioned in my /var/db/pfblockerng/dnsblalias/DNSBL_adverts file, change it to something your blocking..:

#!/usr/local/bin/php -f
error_reporting(E_ERROR | E_PARSE);
global $_SERVER;
$_SERVER['HTTP_HOST'] = "on-mobi.com";
echo "\nTEST-START\n";
for($i=0;$i<$argv[1];$i++){
	if ($i % 100 == 1)
		echo ".";
	include('pfblockerng/www/index.php');
}
echo "\nTEST-END\n";

Make it executable:

chmod +x /root/testpfb.sh

And create a logfile executing it, this 'should' also hang..:

truss -Haedf -s 100 -o /root/truss_pfblocker_test.log /root/testpfb.sh 1

Preferably create 2 logfiles, 1 while webgui is unresponsive, and one when everything still works. That makes it easier to compare the two.. Also when webgui is unresponsive double check that the culprit still looks like it might be coming from pfBlockerNG..
Below command should return a number above 100..

/usr/bin/sockstat | grep lighttpd | wc -l

Or actually it might just be waiting for php-fpm which is 'bussy'…

#####################################################
Other separate request, apply the patch below, restart both php-fpm and webgui from the console options 11 and 16.
https://patch-diff.githubusercontent.com/raw/pfsense/pfsense/pull/3769.patch
Then every once in a while run the following on the pfSense box, preferably before everything is already broken or asap after, as it needs a socket available to itself as well it might not work once the problem is presenting itself..:

curl -k https://localhost/status

This should show some output on what php-fpm is busy with.. And might show if some/what script is taking more time than it should..

bgbird03

Is there an "easy button" for this fix? Maybe just disable pfBlockerNG? I'm pretty uninitiated…just happy I found out "why" it is happening because I had no clue. I'm assuming this will be fixed in 2.4.3 or an update to the extension?

ivor

@pppfsense:

I am very surprised that this was not caught in testing: Many, many people run pfBlockerNG, Suricata/Snort and Squid. That should be a basic configuration to be tested.
Yes, it takes traffic and some time to manifest, but any decent QA dept. needs to have, beyond load producing tools, monitoring tools to watch for memory leaks and process status (I did SW QA a few years ago).

pfBlockerNG wasn't even part of our QA, we were too busy with 11.0 > 11.1 switch. We love and use pfBlockerNG but it's developed by our community member BBcan177 and not pfSense team (even though we fully support him!). In BBcan177's defense, he had very little time for testing with our switch to FreeBSD version 11.1. To learn about what I am talking about see the following blog post: https://www.netgate.com/blog/no-plan-survives-contact-with-the-internet.html

ivor

@bgbird03:

Is there an "easy button" for this fix? Maybe just disable pfBlockerNG? I'm pretty uninitiated…just happy I found out "why" it is happening because I had no clue. I'm assuming this will be fixed in 2.4.3 or an update to the extension?

Disabling pfBlockerNG's DNSBL is a temporary fix until the real cause is found.

chrcoluk

interestingly following me clean installing 2.4.2 using config.xml (so pfblocker NG updated itself also).

Using my FPM patch I had no 502 errors, and 7 FPM processes were running presumably to satisfy all the php scripts running to process everything and to satisfy the webui requests. So I think my patch works well.

Jim has already gave it the thumbs up on a PR, and Martin has committed the code as well.

https://redmine.pfsense.org/issues/8125#change-35234

D-Kun

Hi,

I am currently stable for about 6 days.

2.4.2-RELEASE
pfBlockerNG 2.1.2_1 (updates all lists once a day)
snort 3.2.9.5_3

I was just removing the Status_Traffic_Totals (vnstat) package.
Before I was getting 502.

fortrash

Hello,

since yesterday I haven't had any problem. No hickups or any breakdown.

The script:

truss -Haedf -s 100 -o /root/truss_pfblocker_test.log /root/testpfb.sh 1

-> does not hang

Question:

What have I to apply to fix the issue? The patch that was made available via git or the fixes below:

https://forum.pfsense.org/index.php?topic=110515.60

Have you already commited the changes?

Thanks,

chrcoluk

the git patch commit by Martin is the the same as my patch, he just made it an easier process and we can wait for the pfsense staff to approve the commit so it makes it into a future version, but for now either apply my patch in its raw format or the use the git id Martin provided.

As long as your unit has at least a gig of ram it should solve the problem.

Note the patch was created on a 2.4.x unit, I am not sure how clean it will apply on 2.3.x devices, thats untested.