pkg-static consuming 100% cpu
-
Same here. Just started doing it today. Earlier this week, I had to take action because 'rate' was consuming 100% CPU on both cores on my dual-core VM. Now today, pkg-static was pegging one core until I killed its pid.
-
Any temporary workarounds?
-
I bumped into this accidentally, while trying to install iperf to run some tests.
It actually messed up my /var/db/pkg files and the pkg has been compaining about them since. partially fixed by setting the branch to development and saving then reverting to stable and saving again but not certain it fixed the issue. pkg doesn't complain about not finding files but I see nothing under packages available for install and running pkg update spits errors after about 30 min of being stuck. 1 core pegged to 100% but the system is still operational and routing (hopefully as supposed).
Agree, NOT a great way to fail for ANY process, I would actually go as far as saying not an acceptable way; I'm yet to see any other OS fail to update in this way. -
None I am aware of. Process will return. Wait for the server to come up.
-
I bet all it takes to fix this is a couple of lines at the beginning of a certain file, something along the lines of " ping.......; if/while........; else "try again at another time when server is up and running".
Wondering if this applies to community edition only or paid members are affected as well. -
It's probably even simpler than that, this is a symptom of a common oversight.
-
Just update the certificate and reboot the server. ;-)
-
@artichost said in Pkg-Static 100% CPU:
Just update the certificate and reboot the server. ;-)
Sorry. I'm not shure I understand what you mean. Can you explain please?
-
Kind of a big oversight that a failing update-server(s) can cause pfSense to hog a process at 100% ;)
-
@gonace said in Pkg-Static 100% CPU:
Kind of a big oversight that a failing update-server(s) can cause pfSense to hog a process at 100% ;)
Well the pkg system is part of FreeBSD, feel free to create a bug report at: https://bugs.freebsd.org
-
@grimson said in Pkg-Static 100% CPU:
@gonace said in Pkg-Static 100% CPU:
Kind of a big oversight that a failing update-server(s) can cause pfSense to hog a process at 100% ;)
Well the pkg system is part of FreeBSD, feel free to create a bug report at: https://bugs.freebsd.org
I'm not sure since I've not confirmed it on multiple machines, but I can't reproduce it under FreeBSD 11.2 (clean install) so I'm not sure if it's a bug in FreeBSD rather than something specific when running pfSense.
But yeah, just tried to reproduce it on one virtual machine.
-
@gonace said in Pkg-Static 100% CPU:
I'm not sure since I've not confirmed it on multiple machines, but I can't reproduce it under FreeBSD 11.2 (clean install) so I'm not sure if it's a bug in FreeBSD rather than something specific when running pfSense.
Well the FreeBSD servers aren't down.
-
@grimson said in Pkg-Static 100% CPU:
@gonace said in Pkg-Static 100% CPU:
I'm not sure since I've not confirmed it on multiple machines, but I can't reproduce it under FreeBSD 11.2 (clean install) so I'm not sure if it's a bug in FreeBSD rather than something specific when running pfSense.
Well the FreeBSD servers aren't down.
Well, I blocked traffic to them and even blocked access to the internet on FreeBSD 11.2 (clean install) and did the same with a clean install of pfSense and the problem only arises on the clean install of pfSense.
-
@p1ter It´s a handshake problem ;-)
-
The BSD servers might be up but blocking access to them upstream would be equivalent to them being down. Can someone try to do just that?
Definitely a very disturbing oversight, makes me wonder what other things received the same level of attention :) -
@artichost Thank you. I mean can I fix it by myself?
-
@veriqster said in Pkg-Static 100% CPU:
The BSD servers might be up but blocking access to them upstream would be equivalent to them being down.
That would make them completely unreachable, it currently looks like the pfSense servers hang/are very slow during the SSL negotiation which is a different situation. If you want to see the difference do a traffic capture instead of fooling around.
-
Hello from Brazil. Same here, pkg-static-update 100% usage cpu but the system looks functional.
-
Hello everyone,
I have the same problem but also the process /usr/local/sbin/pkg-static search -r to 100% -
I'm not an expert, but could one via ssh temporarily replace pkg-static with a shell script that basically does nothing? If so, can someone with more skills than myself show me the way?
-Chance -
then maybe someone should just pull the Ethernet plug out of them until things are sorted out instead of messing up with some systems that people depend on.
Honestly I would have said "shit happens, maybe they can just fix it going forward" but his attitude of leaving the servers on knowing it causes issues like this makes me wonder what other things are deemed as acceptable by "the man in charge", things that a dumb user like me might not be able grasp even by looking at the code (which I have neither time nor desire to dedicate myself to) -
Guys, just disable the dashboard automatic update check, the process will disappear after some time. No reboot or anything is needed.
Yeah its annoying to eat one core after the update server broke, but not this dramatic.
I'm sure the Devs will fix this issue in future.-Rico
-
Identical issue with my copy of the community version here (2.4.4-RELEASE-p2, upgraded three times since installation).
I've killed off the pkg-static process with "kill" and disabled the "Dashboard Check" for updates. This seems to bring CPU usage back to normal, provided that I don't open the Update screen.
-
I've done the following and corrected (temporarily) the 100% CPU.
Via ssh, I moved /usr/local/sbin/pkg-static to /usr/local/sbin/pkg-static.old and created /usr/local/sbin/pkg-static (a shell script) with the following contents.
while :
do
sleep 5000
done;don't forget to chmod +x /usr/local/sbin/pkg-static
This has corrected my CPU usage until the servers are back up. I am not sure if this survives a reboot, but as such it's easy to implement.
-
I disabled updates in the dashboard in /system_update_settings.php 30 minutes ago, no change. Did some of you guys had any luck with that?
-
Can take up to ~1 hour. But it will stop, trust me.
-Rico
-
@rico It just did. I guess I didn't wait long enough. Thanks for the tip.
-
@rico Disable automatic update. Nice tip! Thanks from Brazil!
-
@rico said in pkg-static consuming 100% cpu:
Problem disappeared
-
Package servers appear to be up now, at least from my location.
-
Same here, looks like back in business.
last pid: 92723; load averages: 0.70, 0.34, 0.14 up 0+00:02:30 19:40:23 153 processes: 3 running, 103 sleeping, 47 waiting Mem: 51M Active, 45M Inact, 70M Wired, 15M Buf, 286M Free Swap: 102M Total, 102M Free PID USERNAME PRI NICE SIZE RES STATE TIME WCPU COMMAND 11 root 155 ki31 0K 16K RUN 2:01 100.00% [idle] 0 root -16 - 0K 224K swapin 0:26 0.00% [kernel{swapper}] 4 root -16 - 0K 32K - 0:01 0.00% [cam{scanner}] 12 root -60 - 0K 752K WAIT 0:01 0.00% [intr{swi4: clock (0)}] 338 root 52 0 62804K 31748K accept 0:01 0.00% php-fpm: pool nginx (php-fpm) 337 root 30 0 62804K 31876K piperd 0:01 0.00% php-fpm: pool nginx (php-fpm) 339 root 52 0 65112K 33784K accept 0:01 0.00% php-fpm: pool nginx (php-fpm){php-fpm} 4 root -16 - 0K 32K - 0:00 0.00% [cam{doneq0}] 43971 root 20 0 23592K 8824K kqread 0:00 0.00% nginx: worker process (nginx) 25 root 16 - 0K 16K syncer 0:00 0.00% [syncer] 40067 unbound 20 0 23664K 13920K kqread 0:00 0.00% /usr/local/sbin/unbound -c /var/unbound/unb 15 root -68 - 0K 160K - 0:00 0.00% [usb{usbus0}] 0 root -92 - 0K 224K - 0:00 0.00% [kernel{em0 taskq}] 12 root -88 - 0K 752K WAIT 0:00 0.00% [intr{irq17: mpt0 ehci0}] 28768 root 20 0 6900K 2312K nanslp 0:00 0.00% [dpinger{dpinger}] 44956 root 20 0 10352K 10468K select 0:00 0.00% /usr/local/sbin/ntpd -g -c /var/etc/ntpd.co 15 root -68 - 0K 160K - 0:00 0.00% [usb{usbus1}] 78824 root 20 0 6400K 2544K select 0:00 0.00% /usr/sbin/syslogd -s -c -c -l /var/dhcpd/va
-Rico
-
I split this off from the previous thread that was almost a year old. Locking this since it's solved and to prevent similar thread necromancy from happening here.