New 502 Bad Gateway

doktornotor · Sep 26, 2017, 7:51 PM

Well, then you'll need this as well, probably…

BreeOge · Sep 26, 2017, 11:45 PM

Thank you sir. I will try that as well. I thank you for your help.

BreeOge · Sep 30, 2017, 5:07 AM

I have a question Doktornotor, will rc.php_ini_setup get overwritten on updates?

BreeOge · Sep 30, 2017, 5:33 PM

Also its still happening.. even with the changes you suggested :(

BreeOge · Oct 2, 2017, 3:35 AM

Update* Reverted back to 2.4.0 and its no longer happening, only happens with the 2.4.1 version!

BreeOge · Oct 4, 2017, 5:55 PM

Worked great for the 2 days, updated to the OCT 3 update, and now all my sites are getting 502 bad gateway. Trying the new update today.. But something is seriously messed up.

BreeOge · Oct 4, 2017, 7:40 PM

All I get now is upstream timed out, however nothing before this starts, it acts as if everything is ok, then bam 502 error. No Listen queue overflow nothing. something is killing Nginx.

Oct 4 14:18:53	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:53 [error] 34017#100128: *892 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144432540 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:18:43	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:43 [error] 34017#100128: *890 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144422541 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:18:42	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:42 [error] 49658#100106: *381 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.139, server: , request: "GET /decide?version=1&lib=android&token=0350ff19c3e127ef2c92f5cc82284391&distinct_id=d15cf0351e8da34a2092d588fd3e33ed7fecd29e HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "decide.mixpanel.com"
Oct 4 14:18:33	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:33 [error] 34017#100128: *888 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144412542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:18:32	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:32 [error] 49658#100106: *379 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.152, server: , request: "GET /MFEwTzBNMEswSTAJBgUrDgMCGgUABBTBL0V27RVZ7LBduom%2FnYB45SPUEwQU5Z1ZMIJHWMys%2BghUNoZ7OrUETfACEAi4elAbvpzaLRZNPjlRv1U%3D HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "ocsp.digicert.com"
Oct 4 14:18:23	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:23 [error] 34017#100128: *765 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144402542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:18:23	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:23 [error] 34017#100128: *879 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144382542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:18:23	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:23 [error] 34017#100128: *651 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144392541 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:18:14	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:14 [error] 49658#100106: *375 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:18:09	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:09 [error] 49658#100106: *373 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:18:01	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:18:01 [error] 49658#100106: *371 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:17:59	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:17:59 [error] 49658#100106: *369 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:17:59	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:17:59 [error] 49658#100106: *367 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:17:56	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:17:56 [error] 49658#100106: *363 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:17:54	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:17:54 [error] 49658#100106: *360 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:17:54	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:17:54 [error] 49411#100110: *359 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.124, server: , request: "POST /news/report HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "n.m.ksmobile.net"
Oct 4 14:15:53	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:53 [error] 34017#100128: *892 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144372542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:15:43	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:43 [error] 34017#100128: *890 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144362541 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:15:33	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:33 [error] 34017#100128: *888 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144352542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:15:32	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:32 [error] 49658#100106: *335 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 192.168.1.152, server: , request: "GET /GIAG2.crl HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "pki.google.com"
Oct 4 14:15:23	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:23 [error] 34017#100128: *765 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/snort_alerts.widget.php?getNewAlerts=1507144342542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:15:23	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:23 [error] 34017#100128: *879 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "POST /widgets/widgets/gateways.widget.php HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"
Oct 4 14:15:23	fort-smith.rentlynnwood.com		nginx: 2017/10/04 14:15:23 [error] 34017#100128: *651 upstream timed out (60: Operation timed out) while reading response header from upstream, client: 70.178.196.154, server: , request: "GET /widgets/widgets/pfblockerng.widget.php?getNewCounts=1507144342542 HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.socket", host: "fort-smith.rentlynnwood.com", referrer: "https://fort-smith.rentlynnwood.com/index.php"

doktornotor · Oct 4, 2017, 7:45 PM

As noted, you are simply overloading nginx with your CP. All the hints are already above, go on with some better tuning. Noone will do it for you.

BreeOge · Oct 4, 2017, 7:55 PM

I have upped the tuning, all the things you suggested. Also they does not explain the fact that it worked fine for 2 full days, after I reverted to 2.4 then on OCT 3 after the update less than an hour it started happening. Its not overloading the CP, it looks like it is, but its not. It just dies.. you see the CP because when it dies, it tosses the 502 and the CP stops. I have the settings all set way up, to stop this, and it has had no effect at all! Once the Nginx crashes nothing can work so it starts tossing out errors. Also I have turned off the CP just to see if that was the issue, it still happened.

I had PFblocker installed, just removed it to see if that is causing the issue. 30 min, no crash as of yet.

doktornotor · Oct 4, 2017, 8:19 PM

Eh well, it not just overloads nginx, it overloads the entire box apparently as mentioned earlier.


Sep 25 11:25:10	kernel		sonewconn: pcb 0xfffff8004c40f690: Listen queue overflow: 193 already in queue awaiting acceptance (9 occurrences)
[/code)[/code]

BreeOge · Oct 4, 2017, 8:39 PM

Still good to go since I removed PFblocker, I am thinking it may have something to do with PFblocker / Snort/ or Squidguard working together on 2.4 What was updated on 2.4.0 on the 3rd? I looked for patch notes, but could not find it. Any file that shows what was updated on past updates?

When it does lock up, the SSH becomes unresponsive as well, it only shows the basic login, I have to CTRL-Z to close the process.

*** Welcome to pfSense 2.4.0-RC (amd64) on fort-smith ***
^ZYou have stopped jobs.
[1] + Suspended /etc/rc.initial

reboot

That is the only way to get it to free up. If you try to restart PHP it will just lock up as well.

I also updated the listen Queue overflow, and the error for overflow does not happen now till WAY later, if its been locked up for 15 min or more. That is a symptom. But some file that changed on the 3rd is causing issues between pfblocker, snort, or squidguard working with each other.

Still working good since I removed Pfblocker, no errors what so ever in the logs, running like a champ.

doktornotor · Oct 4, 2017, 8:55 PM

If you read the above message, your box is completely overloaded with queued ingress connections. So yeah, in that state NOTHING works properly. Even linked you to the kernel tunable.

Also: http://nginx.org/en/docs/freebsd_tuning.html

BreeOge · Oct 4, 2017, 9:26 PM

I really appreciate the help doktornotor, and I understand where your coming from and why your giving me the solutions you are suggesting. I guess I have not explained it well enough though. I understand that its being queued with ingress connections, but the problem I have is why does it only happen when I updated on the 3rd to the newest 2.4.0 version and not before, system load has not changed, actually if anything system load is currently low. Problem only showed up on 2.4.1 and on 2.4.0 after Tuesdays update. Before that I never had an issue with Overloaded ingress connections even with pfblocker installed. I have had the system with a heaver load than it currently has and never received these errors. I have tried the settings you suggested mine are currently

kern.ipc.somaxconn=4096

and

backlog= 4096

and I still have the issue.

However when I removed pfblocker the issue stopped, but its never been an issue before, and like I said its actually low load at the current time. It only raised its head after the 2.4.0 update on the 3rd or on the 2.4.1 updates. Before then even on the 2.4.0 I had no issues even during high loads. That is why I posted in the 2.4 Development Snapshots, because it only showed up after an update, something changed to cause the problem. All I did was update to newest version, no load change or anything else. Update then issue showed up. Again I do appreciate your help, but I guess I will have to wait till more are having the issue. I know they will have issues, I have 2 boxes set up, one only has one to two users at any given time so super low load and even it started after the update.

Thank you for your assistance. Sorry for taking your time.

BreeOge

seanr22a · Oct 5, 2017, 8:07 PM

I'm seeing the same thing - 502 errors. I'm running three sites all of them worked perfect until 2.4.0-RC changed from bsd 11.0 to bsd 11.1 (a few days ago)
Strange thing it's only one site that have problems. The problem site is the site with most users/traffic.

This morning I removed pfBlocker on the problem site and so far everything good. It's to much of a coincidence that the problem started when the move was made from bsd 11 to 11.1 a couple of days ago.

I had to remove squid on all three sites (XMLRPC Sync stopped working giving a constant flow of errors) and now Pfblocker is removed on one site after the introduktion of bsd 11.1.

BreeOge · Oct 6, 2017, 4:06 AM

Yea, it happens more on the one that has higher traffic, but if you leave it even on the low traffic ones they will eventually get it as well, just takes longer. Its like its not releasing something, and keeps building till it locks up..

I removed all my PFblockers too.. Interesting side note as well

I have 3 systems, 2 of them have Pfblocker, squid, and squidguard installed. Those are the ones getting the 502. I have another one set up, with only Squid, and Pfblocker, and Squidguard disabled. It has yet to get a 502 Bad gateway. It seems to be something with the 11.1 and either Pfblocker or Squidgard or all working together, but if you have Squid and Squidguard and no pfblocker it does not happen, same goes with Squid and Pfblocker with no Squidguard. Its only when all 3 are present.

Seanr22a do you have those 3 installed on all your systems or a mix?

seanr22a · Oct 6, 2017, 5:28 AM

@BreeOge:

Seanr22a do you have those 3 installed on all your systems or a mix?

I have two SG-2440 and one SG-8860. They all have Snort, Squid and Pfblocker. The configuration is identical at these sites so I use XMLRPC Sync in all three packages.
At the two sites still running Pfblocker I have moved the IP blocklists to Snort and only using DNSbl in the Pfblocker package. I will wait a couple of days if its still fine I will install Pfblocker using only DNSbl at the main site as well and give it a try. My problem with squid is the XMLRPC Sync but all this started to happen with bsd 11.1

To me it looks like it's running out of resources of some kind. I have a lot memory and diskspace available when it happens so its something else.

BreeOge · Oct 6, 2017, 2:07 PM

I have a lot memory and diskspace available when it happens so its something else.

Same here..

The release notes for 11.1 are here https://www.freebsd.org/releases/11.1R/relnotes.html#kernel

Lots of changed for the kernel. I suspect one of these are the cause, due to it crashed the kernel but I could be wrong on that. Also I have no idea what would cause it. just something to look at.

I know it looks like the system is being overloaded like doktornotor said, but hard to believe a system that never overloaded before the update now overloads constantly with the same load or less. Does not make sense.

luckman212 · Oct 6, 2017, 2:11 PM

Hmm I'm not running into any issues on my test 2.4.1 box, but to be fair, I'm not running Snort of pfBNG. Hope this doesn't turn out to be a nasty bug in the kernel that could delay 2.4.x getting out the door. Seems like every time 2.4 is about to drop something else comes up. Been 542 days since 2.3-RELEASE :)

BreeOge · Oct 6, 2017, 2:23 PM

@luckman212:

Hmm I'm not running into any issues on my test 2.4.1 box, but to be fair, I'm not running Snort of pfBNG. Hope this doesn't turn out to be a nasty bug in the kernel that could delay 2.4.x getting out the door. Seems like every time 2.4 is about to drop something else comes up. Been 542 days since 2.3-RELEASE :)

Yea it is for sure related to the 11.1 Kernal, PfblockerNG, or Squidguard. Remove one or the other (pfblockerNG, or SquidGuard) and issue does not exist.

Did a test, I had one system working for over 24 hours, since I removed PfblockeNG and no 502, put it back in and in less than 2 hours, crash.

I have one system with PfblockerNG installed, and squidguard disabled, no issues at all. Enabled Squidguard, and crash in less than 2 hours.

So the issue lies between PfblockerNG, Squidguard and the Kernal. I would suspect, but i am no expert.

MaxPF · Oct 7, 2017, 3:24 PM

Started having the same issue since the 2.4RC switch from 11.0 to 11.1. I am running pfBlocker.