LCDProc 0.5.4-dev

tix

2. The LCDd process can start to use a lot of CPU time causing the box to become unresponsive (but still routes traffic OK).

I have just acquired an x700 from work and trying to get it running smoothly and am encountering this problem. It is a clean full install (on HDD) of 2.0.1 with IPv6 git sync'd as of 21 Jan.
I keep running into this within 16-24 hours uptime where the system is non-responsive but still routes although more slowly than normal. I cannot SSH use the webif, or console when this happens. The RRD graphs stop as well. I have only the load and states screens enabled thinking maybe the number of screens was an issue, but today when it became unresponsive it was stuck on the load screen and the current load was displayed as a 24.4!!!

I am also of the (pretty much worthless) opinion that it is too many instances of the LCDProc client as I can see the log entries posted above. I am going to disable LCDProc-dev for now and see if everything stays stable for at least 24 hours.

marcelloc

Also check states table size and usage on Dashboard.

stephenw10

@tix:

I am going to disable LCDProc-dev for now and see if everything stays stable for at least 24 hours.

The more people running this and giving feedback the faster it's likely to be solved. So thanks. :)
If you are running it again try to SSH in. After you enter the password it appears to have frozen but if you wait long enough it will log in (several minutes). Then you can see what's running. E.g.


[2.0.1-RELEASE][root@pfsense.fire.box]/root(4): ps aux | grep lcd
root   16576  0.0  0.3  3656  1508  ??  S    10:59PM   0:00.00 /bin/sh /tmp/lcdclient.sh
root   16628  0.0  3.4 47452 17388  ??  S    10:59PM   0:00.31 /usr/local/bin/php -f /usr/local/pkg/lcdproc_client.php
[2.0.1-RELEASE][root@pfsense.fire.box]/root(5): ps aux | grep LCD
nobody 16562  0.0  0.3  3368  1464  ??  Ss   10:59PM   0:00.02 /usr/local/sbin/LCDd -c /usr/local/etc/LCDd.conf

You should see one instance of each of those three programs running. If you have more then it has started or stopped incorrectly.

You can also run```
top -SH

LCDd should be consuming 1% or less cpu.

Steve

m4f1050

Here is my recent log of LCDproc after a fresh reboot of my Firebox, right now I have System Time and System Stats working.

EDIT: I had my WAN cable unplugged. I replaced my X500 with my X700 while I did this…


Jan 22 20:37:22 LCDd: Connect from host 127.0.0.1:21770 on socket 13
Jan 22 20:37:22 LCDd: Connect from host 127.0.0.1:34709 on socket 12
Jan 22 20:37:20 LCDd: Listening for queries on 127.0.0.1:13666
Jan 22 20:37:20 LCDd: Using Configuration File: /usr/local/etc/LCDd.conf
Jan 22 20:37:20 LCDd: LCDd version 0.5.5 starting
Jan 22 20:37:17 php: lcdproc: Sync: End package sync
Jan 22 20:37:17 php: lcdproc: Sync: Begin package sync
Jan 22 20:37:17 php: lcdproc: Sync: End package sync
Jan 22 20:37:17 php: lcdproc: Sync: Begin package sync
Jan 22 20:37:17 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)
Jan 22 20:37:17 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)
Jan 22 20:37:17 php: : Restarting/Starting all packages.
Jan 22 20:37:11 check_reload_status: Starting packages
Jan 22 20:37:11 php: : pfSense package system has detected an ip change 0.0.0.0 -> ... Restarting packages.
Jan 22 20:37:11 php: : OpenNTPD is starting up.
Jan 22 20:37:05 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)
Jan 22 20:37:05 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)
Jan 22 20:36:57 php: : ERROR! PPTP enabled but could not resolve the $pptpdtarget
Jan 22 20:36:56 php: : The command '/usr/bin/killall 'ntpd'' returned exit code '1', the output was 'killall: warning: kill -TERM 59839: No such process'
Jan 22 20:36:56 php: : Creating rrd update script
Jan 22 20:36:56 php: : Resyncing OpenVPN instances for interface LAN.
Jan 22 20:36:54 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)
Jan 22 20:36:54 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)

tix

@stephenw10:

The more people running this and giving feedback the faster it's likely to be solved. So thanks. :)
If you are running it again try to SSH in. After you enter the password it appears to have frozen but if you wait long enough it will log in (several minutes). Then you can see what's running. E.g.

You should see one instance of each of those three programs running. If you have more then it has started or stopped incorrectly.

You can also run```
top -SH
LCDd should be consuming 1% or less cpu.

Steve

Thanks Steve but I've tried that and even after more than an hour I still cannot get the menu so I usually give up and do a power-cycle as that is the only way I've found to recover.
I would love to be able to confirm if there are multiple instances as well as memory usage and such but so far that is not an option.

I forgot to mention previously that someone had suggested that when the connection drops that all the packages are restarted so I disabled the gateway monitor but that made no difference.

The next time I get to watch it for a full day (next weekend?), I will keep an SSH session up running top and hopefully capture some useful information.

EDIT:
One more thing: The looking at the RRD graphs (processor, memory, & states), there is nothing indicating an increase in processes or utilization. Everything is "normal" and then just stops graphing until I reboot.

m4f1050

Here is the X700's System Log pertaining LCDproc 0.5.5 v0.8


Jan 22 22:26:33 LCDd: Connect from host 127.0.0.1:42443 on socket 13
Jan 22 22:26:31 LCDd: Connect from host 127.0.0.1:61742 on socket 12
Jan 22 22:26:29 LCDd: Listening for queries on 127.0.0.1:13666
Jan 22 22:26:29 LCDd: Using Configuration File: /usr/local/etc/LCDd.conf
Jan 22 22:26:29 LCDd: LCDd version 0.5.5 starting
Jan 22 22:26:26 php: lcdproc: Sync: End package sync
Jan 22 22:26:26 php: lcdproc: Sync: Begin package sync
Jan 22 22:26:26 php: lcdproc: Sync: End package sync
Jan 22 22:26:26 php: lcdproc: Sync: Begin package sync
Jan 22 22:26:26 php: : Restarting/Starting all packages.
Jan 22 22:26:21 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)
Jan 22 22:26:20 check_reload_status: Starting packages
Jan 22 22:26:20 php: lcdproc: Failed to connect to LCDd process Operation timed out (60)

top -S for LCDd


PID.....USERNAME.THR.PRI.NICE.SIZE...RES....STATE.TIME.WCPU.COMMAND
27586.nobody......1.....74..r30...3368K.1488K.RUN.....0:00.0.00%.LCDd

tix

I turned LCDd on overnight and left my SSH sessions active running top -SH
This morning I have an unresponsive machine except the SSH sessions are still active. (Also, everyone was asleep so there was no usage from LAN to WAN.)

I killed the LCDd process after capturing the information below and load drops to less than .10 but I cannot restart it. If I use the webif it will generate the previously mentioned log entries and the same occurs starting it from the CLI.

top -SH

last pid: 31188;  load averages:  7.99,  7.75,  6.91        up 0+17:13:06  08:55:32
92 processes:  11 running, 68 sleeping, 1 zombie, 12 waiting
CPU: 18.9% user,  0.0% nice, 80.6% system,  0.5% interrupt,  0.0% idle
Mem: 62M Active, 15M Inact, 39M Wired, 34M Buf, 118M Free
Swap: 512M Total, 512M Free

  PID USERNAME PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
34608 nobody    74  r30  3368K  1488K RUN     77:11 100.00% LCDd
   10 root     171 ki31     0K     8K RUN    797:50  0.00% idle
   11 root     -28    -     0K    96K WAIT     1:57  0.00% {swi5: +}
45412 root      44    0  6588K  4776K kqread   0:56  0.00% lighttpd
   11 root     -32    -     0K    96K WAIT     0:50  0.00% {swi4: clock}
    0 root     -16    0     0K    48K sched    0:44  0.00% {swapper}
34700 root      76    0 33092K 16244K RUN      0:31  0.00% php
45906 root      44    0  3712K  1960K RUN      0:31  0.00% top
32600 root      76    0 33092K 16136K RUN      0:31  0.00% php
40548 root      76   20  3656K  1492K piperd   0:14  0.00% sh
   13 root     -16    -     0K     8K -        0:13  0.00% yarrow
24126 root      44    0  4948K  2516K select   0:12  0.00% syslogd
19804 root      44    0  7992K  3604K select   0:11  0.00% sshd
   19 root      44    -     0K     8K syncer   0:07  0.00% syncer
35740 root      48    0 37188K 18800K piperd   0:07  0.00% php
21686 root      44    0  3316K   924K piperd   0:06  0.00% logger

System.log tail

 tail /var/log/system.log
Jan 23 02:35:21 pfsense dnsmasq[6805]: possible DNS-rebind attack detected: appsforbb.com
Jan 23 02:58:07 pfsense dnsmasq[6805]: read /etc/hosts - 31 addresses
Jan 23 03:26:29 pfsense dnsmasq[6805]: read /etc/hosts - 31 addresses
Jan 23 03:58:37 pfsense dnsmasq[6805]: read /etc/hosts - 31 addresses
Jan 23 05:34:21 pfsense dnsmasq[6805]: read /etc/hosts - 31 addresses
Jan 23 06:34:48 pfsense dnsmasq[6805]: read /etc/hosts - 31 addresses
Jan 23 07:35:19 pfsense dnsmasq[6805]: read /etc/hosts - 31 addresses
Jan 23 07:59:59 pfsense dhclient: RENEW
Jan 23 07:59:59 pfsense dhclient: Creating resolv.conf

stephenw10

Further confirmation of a problem. Thanks.
By the way pfSense uses circular logs so you have to use clog to tail them.
See: http://doc.pfsense.org/index.php/LOGS:Why_can%27t_I_view_view_log_files_with_cat/grep/etc%3F%28clog%29

Steve

m4f1050

Well, today when I woke up and saw my Watchguard I noticed the "Thanks for using pfSense" and hung LCDd… I can restart it, but it would be nice for it not to crash. Will keep testing... All I did was add Interface (and on the dropdown kept WAN)

mdima

Hello everybody,
I just updated the package. The news are the following:

Fix the uptime screen
In the settings, "Enable LCDproc" becomes "Enable LCDproc at startup"
Added "After Install Info" message
Limited the client loop to three times. After three errors connecting to the LCDd service the client will end

I hope the 4th point will solve the problems for the Firebox boxes! Just wait for the 0.9 version of the package and update it…

Thanks to all for the feedbacks,
Michele

EDIT: With a feedback, please post the driver/port you're using, the values of the options of the service and the list of the screens active, thanks!!

power_matz

Hello,

can I leave the LED backlight on all the time? Is this implemented in this update?
Or will it come in a future version?

M.

stephenw10

@power_matz: The backlight control is part of the driver not the package so this won't have changed.

@mdima: Awesome work! ;D I can't wait to test it.

Steve

fmertz

@power_matz:

can I leave the LED backlight on all the time? Is this implemented in this update?

We are currently focusing on fixing the issue of multiple instances of the server, as well as building the code for driving the LEDs. Once that is stable, we can revisit the user parameters.

tix

Well the update didn't help for me… :(


Mem: 44M Active, 15M Inact, 37M Wired, 28M Buf, 139M Free
Swap: 512M Total, 512M Free

48731 nobody    74  r30  3368K  1488K RUN     77:11 100.00% LCDd

Running clog of /var/log/system.log showed no entries related to the LCD processes.

However it DID continue to work properly this time although the screens changed about every 7 secs or so instead of 5 secs with the webif non-responsive and new SSH sessions not completing to connect. Existing SSH sessions continue just slowly as expected under the load.

For reference all my settings for the Firebox x700 are "default" with "2x20 columns" and the SDEC driver on /dev/lpt0. The only screens enabled are: Uptime, Load, States, Mbuf, & Interface Traffic (WAN).
GW monitoring is disabled.

EDIT: Forgot to add

like the new Uptime screen! Much cleaner.
in my set up, it seems to be failing after around 16-18 hours uptime and/or around 4-5am, not sure which yet, with no network activity occurring.

mdima

@tix:

For reference all my settings for the Firebox x700 are "default" with "2x20 columns" and the SDEC driver on /dev/lpt0. The only screens enabled are: Uptime, Load, States, Mbuf, & Interface Traffic (WAN).
GW monitoring is disabled.

Hi Tix,
thanks for the info! I am now running my secondary machine with the same screens in order to test the "client" part of the package. If there is something wrong in the client for this screens I will find out soon. Of course I can only test the client because I use a different driver… but if the test goes OK it means some problem in LCDd 0.5.5 or the driver.

@tix:

EDIT: Forgot to add

like the new Uptime screen! Much cleaner.

in my set up, it seems to be failing after around 16-18 hours uptime and/or around 4-5am, not sure which yet, with no network activity occurring.

Thanks for apreciating! I will see tomorrow how my secondary machine is doing…

Ciao,
Michele

m4f1050

You think you can program more stuff on the LCDproc MENU?

Maybe stuff like webGUI restart, backup config, restore config, reboot system

Thanks! Doing an awesome job!

To all, I am using GWXepc for the Arm/Disarm LED to turn it green when pfSense comes up and red when it goes down.
What I did was edit the beep.sh and in "start" I added GWXepc -l green and in "stop" I added GWXepc -l red

:)

stephenw10

I've now been running 0.54 versions of LCDd and sdeclcd for 24hrs and am still experiencing the same LCDd CPU usage lock out. So I think we can rule out going to 0.55 as a problem.
Interestingly I was able to observe it happening this afternoon and it ramps up slowly as if it's looping around creating steadily more and more processes until it hits 100% cpu.

One way to test this would be to compile the old driver against 0.55 and run that. I don't have a suitable compile environment setup at the moment though.

Steve

mdima

Steve,
how many instances of the client were running at that time? Do you have any log that evidence problems or concurrent clients running in the same time?

Thanks,
Michele

stephenw10

Hmm, I'm not sure what you mean.
There were two distinct problems I experienced.
Firstly where multiple copies of the client ended up running. This happened immediately after either rebooting or restarting the service.
Secondly where LCDd ends up using 100% cpu. Just one client can be running when this happens.
I don't have much by way of logging. Can we increase the logging level in lcdd.conf?

Steve

mdima

@stephenw10:

Hmm, I'm not sure what you mean.
There were two distinct problems I experienced.
Firstly where multiple copies of the client ended up running. This happened immediately after either rebooting or restarting the service.

Ok, but thanks to the last change the "second client" was stopped after 2 more attempts to connect, I guess.

@stephenw10:

Secondly where LCDd ends up using 100% cpu. Just one client can be running when this happens.
I don't have much by way of logging. Can we increase the logging level in lcdd.conf?

Ok, that's awesome. At least we know that the issue is not related to multiple instances of the client running, which in any case should not give any problem (since LCDd is made to support different clients in the same time), but for sure it is not a "resoruce leak" because of the PHP client bothering the system…

Thanks,
Michele