Hard Crashing - Out of Memory



  • Hi,

    I rearranged my rules today to sort out a few other problems I was having and also to clear out some of the 6 year old crud in my rulesets. Since then I've been getting the following errors first "randomly" and now every time I click on my OpenVPN tab in Rules.

    The firewall is a VM and it has 6GB ram allocated to it of which it's currently sitting at about 30% (so 2GB). I'm guessing I need to set a larger buffer somewhere but I've been through every forum post and setting on the firewall and nothing has made a difference. The error is:

    Crash report begins.  Anonymous machine information:
    
    amd64
    11.2-RELEASE-p6
    FreeBSD 11.2-RELEASE-p6 #3 518496b29ae(RELENG_2_4_4): Wed Dec 12 07:41:44 EST 2018     root@buildbot2.nyi.netgate.com:/build/ce-crossbuild-244/obj/amd64/ZfGpH5cd/build/ce-crossbuild-244/pfSense/tmp/FreeBSD-src/sys/pfSense
    
    Crash report details:
    
    PHP Errors:
    [13-Mar-2019 18:43:03 Europe/London] PHP Fatal error:  Allowed memory size of 536870912 bytes exhausted (tried to allocate 375427072 bytes) in /usr/local/www/firewall_rules.php on line 88
    [13-Mar-2019 18:43:03 Europe/London] PHP Fatal error:  Allowed memory size of 536870912 bytes exhausted (tried to allocate 375386112 bytes) in Unknown on line 0
    

    Anyone have any idea why it's doing this and/or what I need to change to allocate more RAM to whatever process is running out of memory?

    G



  • My instance runs several packages and it works just fine with 2GB RAM total, of which 26% is used. You have some sort of problem.

    Which packages, if any, are you running?



  • Thanks but nothing exciting.

    ACME which I've never used
    Lightsquid which is configured and working
    Open-VM-Tools
    pfBlockerNG set up with most of the blocklists (only about 20 lists and none of them more than half a million IP's)
    snort set up with a relatively small selection of rules (I'd say about half)
    squid set up as a transparent proxy
    System_Patches which I've never used

    But as I said it's not "out of memory" on the main status page, far from it. Just don't know where to look or what to tweak.

    G



  • You have pretty much every heavy package installed. Squid can suck up amazing gobs of RAM all by itself. Squid is practically useless for caching these days, so unless you're using it for URL filtering with squidguard, I would remove it. If you definitely need squid, what do you have set for its Memory Cache Size?



  • Yeh I have been considering removing Squid for that reason.

    That said, as I mentioned I have 6GB allocated to the FW and it's hovering between 30-40% according to the status page. I've not yet seen it go over 3GB used so it's barely touching the sides.

    I should mention this is just for my home system, so it's just 3 people hitting it on a 80Mbit broadband line so it's not exactly getting hammered. It also has shared use of a Intel I7 2600 processor which sits at about 45% utilization. I don't think the resources available to the VM are the issue.

    The Memory Cache Size is set to 1024 so well below the 50% maximum size. Unless these applications are using RAM that isn't reported on the Dashboard then it's no where near limiting.

    Thanks again for having a try though.

    G


  • Netgate Administrator

    @xxgbhxx said in Hard Crashing - Out of Memory:

    536870912

    It's exhausting the 512MB limit for php processes. It appears to be trying to allocate 375MB:

    	printf("<a href=\"diag_dump_states.php?ruleid=%s\" data-toggle=\"popover\" data-trigger=\"hover focus\" title=\"%s\" ",
    	    $rulesid, gettext("States details"));
    

    So when you are mousing over something in the ruleset? Do you have huge aliases? A very large number of states?

    Steve



  • I don't have a large number of states - I've not noticed it goes over 1000 total for the entire firewall. I'm not sure what you'd call huge aliases but within Pf none of the aliases are over 20 IP's. In PfBlockerNG I've got a number (17) of blocklists defined and the largest one when I do a refresh is about 80,000 IP's so tiny on the scale of lists.

    I've just gone and cycled through the interfaces it seems to be my OpenVPN interface that's causing the issue. Even when I just click on it to browse to it that's when it locks up and it's now giving me a 504 error. Thing is I can't get into the GUI to see what rules are there that might be causing the problems. How can I get to the rules and disable/delete to see if that solves the problem?

    G



  • Ok this time now that it's re-started I get

    PHP errors
    
        PHP ERROR: Type: 1, File: /usr/local/www/firewall_rules.php, Line: 88, Message: Allowed memory size of 536870912 bytes exhausted (tried to allocate 375427072 bytes) @ 2019-03-13 18:43:03
        PHP ERROR: Type: 1, File: /usr/local/www/firewall_rules.php, Line: 491, Message: Allowed memory size of 536870912 bytes exhausted (tried to allocate 532692992 bytes) @ 2019-03-13 20:31:43
    
    General
    
        Unresolvable source alias 'Internal_SUbnets' for rule 'OpenVPN wizard' @ 2019-03-13 23:00:09
        Unresolvable source alias 'Internal_SUbnets' for rule 'OpenVPN wizard' @ 2019-03-13 23:00:46
    

    So it's clearly those last two lines. As I mentioned when I click on the tab it crashed the firewall so how do I get in there to delete the offending rule. The only range under "Internal_Subnets" is 172.17.10.0/24 so I don't understand why it's causing these problems.

    Any ideas?

    G



  • Hi,

    @stephenw10 any ideas?

    Thanks

    Gary



  • @xxgbhxx said in Hard Crashing - Out of Memory:

    The only range under "Internal_Subnets" is 172.17.10.0/24 so I don't understand why it's causing these problems.

    This particular alias is just the drip that overflows the bucket.

    A radical test to check if you are on the right track : tell PHP it can use more memory. Do this by changing the php.ini file.

    Edit : etc/rc.php_ini_setup - line 181 - add " memory_limit = 256M "

    ....
    ; File generated from /etc/rc.php_ini_setup
    memory_limit = 256M
    ...
    

    and restart PHP using console option 16.

    Now test.



  • @gertjan said in Hard Crashing - Out of Memory:

    etc/rc.php_ini_setup

    Hi thanks for that.

    So I tried 256M and it had no effect. I increased it to 1024M restarting PHP with option 16 both times and it had no effect.

    PHP ERROR: Type: 1, File: /usr/local/www/firewall_rules.php, Line: 708, Message: Allowed memory size of 536870912 bytes exhausted (tried to allocate 375398400 bytes) @ 2019-03-14 14:04:56
    

    It doesn't seem to increase the allowed memory size at all either way.

    Any other ideas?

    G



  • Place this 'code' in a file called whatever.php into the GUI webroot (/usr/local/www/) :

    <?php
    phpinfo();
    ?>
    

    and open the file like
    http://pfsense.yourlan.tld/whatever.php

    Check the several occurrences of 'memory' : I can't find 512M (!!) or 536870912 bytes ....
    "memory_limit" is the value that should have been changed. With 6 Gb you could even make that value bigger.

    Btw :

    @xxgbhxx said in Hard Crashing - Out of Memory:

    (only about 20 lists and none of them more than half a million IP's)

    Serious ??
    You really ask unbound, the DNS server, to restart with lists close to half a million of IP's ??
    Woooow.



  • @gertjan

    memory_limit = 1024M for both the local and master value

    Not sure what your reference to DNS means. The lists are all IP's so wouldn't go anywhere near DNS.

    Thanks again though

    G



  • He might have meant pfBlocker, not Unbound.



  • @kom said in Hard Crashing - Out of Memory:

    He might have meant pfBlocker, not Unbound.

    Or both.
    pfBlocker populate lists parsed by unbound.

    (I guess it's time that I replay with pfBlocker - but don't do what to block, or who ...)


  • Netgate Administrator

    Or more likely that pfBlocker can add lists to both but DNSBL does not have to be used.

    I believe this is the only line that would need changing:
    https://github.com/pfsense/pfsense/blob/64d2dd619973fa986389de2b5db43b0043c5ddd6/src/etc/inc/config.inc#L43
    But I have not tried it and you should only do so as a test. It should not be trying to allocate that ammount of RAM to open the page.

    Steve



  • @kom

    Ah

    pfBlocker handles lists many millions of IP's in size - well according to the 100's of posts I've read. Many people use much larger IP lists than I do.

    Either way it's not going anywhere near Unbound as the lists I use are all IP address so there is no resolution to be done unless I'm missing something obvious.

    Either way the firewall is no where near ram, disk or CPU limits.

    G



  • @stephenw10

    Do I need to do a firewall re-start for it to pick up the change?

    G


  • Netgate Administrator

    You would likely need to restart php and the webgui from the console menu at least. It might require a reboot.

    Steve



  • @stephenw10

    Increased that value to 2048 and restarted PHP/WEB. Also rebooted. Hasn't allowed me to get into that tab in the GUI as I still get a 504 error however it has now stopped giving me out of memory errors.

    One step forwards but I sense no nearer!

    Thanks again.

    Any other ideas?

    G


  • Netgate Administrator

    2048 is huge, is it exhausting the system RAM?

    It probably isn't because I would expect you might hit this before that:
    https://github.com/pfsense/pfsense/blob/RELENG_2_4_4/src/etc/rc.php_ini_setup#L271

    Check /tmp/php_errors.txt

    Check the main system log.

    I assume restarting php and the wbegui from the menu does not allow you back in?

    Steve



  • @stephenw10

    I did it to ensure that exhausting memory wasn't the problem - appreciate it's way more than needed. I have plenty of system RAM (6GB) allocated to the VM so 2GB isn't anywhere near the limit for the system.

    I'm certain it's something that's corrupted with the OpenVPN rule I have as it's referencing a table that doesn't exist (assuming it's case sensitive)

    php_errors is a zero byte file.

    Main system log looks to be corrupted I think but there are errors noted there from 3 days ago. Can't see anything more recent though.

    Nothing I do allows me to click on the OpenVPN tab under "Rules". The rest of the GUI works fine and does not crash.

    I will have a fiddle.

    G



  • @xxgbhxx said in Hard Crashing - Out of Memory:

    Main system log looks to be corrupted I think but there are errors noted there from 3 days ago. Can't see anything more recent though.

    What are you doing to check this ?



  • @gertjan @stephenw10

    I'm just catting the file like I would any other log file so

    cat /var/log/system.log
    

    I've renamed it to system.log.old and restarted. Log is now logging fine again.

    The only error/issue I can see is

    Mar 14 18:20:20 pfsense2 php-fpm[340]: /rc.filter_configure_sync: New alert found: Unresolvable source alias 'Internal_SUbnets' for rule 'OpenVPN  wizard'
    

    Which is the same error I get if I click on the OpenVPN tab in the GUI (which then goes on to crash the FW)

    I need some way of directly editing the rules for that tab to remove this reference (assuming that's where it is) so that the GUI stops crashing.

    Apart from not being able to get into this part of the GUI the FW seems to be performing as I'd expect.

    G



  • @xxgbhxx said in Hard Crashing - Out of Memory:

    I'm just catting the file like I would any other log file so
    cat /var/log/system.log

    Those files are circular files, the have to be read with

     clog /var/log/system.log
    

    They are managed with Status / System Logs / Settings. The drawback is that resetting will reset all log files.


  • Netgate Administrator

    Check the ruleset in /tmp/rules.debug.

    If you made a change that that you think is causing this you can roll back to a previous config from the console menu.

    Steve



  • @stephenw10

    OK we have progress!

    There are probably 500+ lines in there that say

    pass  in log  quick  on $OpenVPN inet from $Internal_Subnets to any tracker 1424387567 keep state  label "USER_RULE: OpenVPN  wizard"
    

    That is clearly the issue.

    I have NO idea where they've come from - I've not run the OpenVPN wizard as far as I can remember. That said of course I might have and not remembered though why would it have just randomly created 100's of entries in that ruleset?

    If I did run it I have absolutely no way of knowing when. Is there absolutely no way I can just remove the offending lines from the OpenVPN rules from the command line?

    G



  • Ok, then is becomes more clear.

    @xxgbhxx said in Hard Crashing - Out of Memory:

    I've just gone and cycled through the interfaces it seems to be my OpenVPN interface that's causing the issue. Even when I just click on it to browse to it that's when it locks up and it's now giving me a 504 error.

    There is something with the rules on that interface. The GUI throws 504.
    When rules are generated, PHP runs out of memory.

    Export a backup of your config, and remove all firewall rules. Import back in again, rebuild your rules.



  • @gertjan

    OK sorted.

    Thank you very much.

    When I exported the rules it created a 12MB 500,000+ line file. What was extraordinarily bizarre was that when I saved the file from the FW gui, the OS couldn't see the file so I couldn't open it to edit it. Even more bizarre was that if I went to save in the GUI, the FW save dialog could see the files but the OS and any other app couldn't.

    In the end I opened the file "live" into Notepad++ but even then I couldn't actually edit the file. I had to save the file in Notepad++ and then re-open it to actually be able to edit it. First time I've ever had that.

    Anyway, I deleted over 500,000 lines of config added by the OpenVPN Wizard and that took the file from 12MB and 550,000 lines to 61k and 2700 lines. Reloaded that back into the FW and with a bit more tidying up everything is working again and not crashing.

    Thank you very much to everyone who chipped in. Was a "simple" fix I really should have come up with myself in the end but thanks for the help. If nothing else it's massively increased my knowledge.

    G


  • Netgate Administrator

    Wow! Never seen anything even remotely like that. Weird.

    I also noticed there are two aliases there Internal_SUbnets and Internal_Subnets. That seems suspicious, you might want to just check your config file now to see if both are still present.

    Steve



  • @stephenw10

    Hi,

    I saw that too and checked for "SU" and "Su" and only "Internal_Subnets " exists so I have no idea where that came from either.

    I clearly must have accidentally clicked on on the Wizard at some point in the last few days and not noticed leading it to get very confused.

    It all seems good now though.

    Thanks again for your help

    G


Log in to reply