[Solved!] netatalk cause CPU 100%

[bluephoenix]

After I installed netatalk to provide the shared folder for my mac, I got CPU 100% as long as I start my computer.

I checked the afpd.log and I get the following error message:
# tail /var/log/afpd.log
Mar 02 22:20:24.488166 cnid_metad[1674] {usockfd.c:169} (E:CNID): error in accept: Socket operation on non-socket
Mar 02 22:20:24.488171 cnid_metad[1674] {usockfd.c:169} (E:CNID): error in accept: Socket operation on non-socket
message repeated 999 times
Mar 02 22:20:24.488898 cnid_metad[1674] {usockfd.c:169} (E:CNID): error in accept: Socket operation on non-socket
Mar 02 22:20:24.488909 cnid_metad[1674] {usockfd.c:169} (E:CNID): error in accept: Socket operation on non-socket
Mar 02 22:20:24.488915 cnid_metad[1674] {usockfd.c:169} (E:CNID): error in accept: Socket operation on non-socket
Mar 02 22:20:24.488921 cnid_metad[1674] {usockfd.c:169} (E:CNID): error in accept: Socket operation on non-socket
message repeated 999 times
Mar 02 22:20:24.489574 afpd[1673] {main.c:102} (N:AFPDaemon): AFP Server shutting down
Mar 02 22:20:24.489889 netatalk[1671] {netatalk.c:340} (N:AFPDaemon): Netatalk AFP server exiting

In deed, you can find the sub process of netatalk - cnid_metad creates hundreds or thousands error logs per second in "afpd.log" so that I have to stop the service or the log file will fill out my whole hard disk.

my /etc/afp.conf is as follows:
[Global]
hostname = BPLinuxDesk
log file = /var/log/afpd.log

[Homes]
path = Downloads
basedir regex = /home

But one strange thing is this phenomenon will only happen when I auto-start netatalk by using "systemctl enable netatalk". If I disable this service from boot up (systemd), instead I start it after the boot up by manually command input, this phenomenon will not happen any more and everything seems normal. Whatever there is error syslog or not, this computer can share its folder for my macbook. So I strongly doubt this issue is caused by systemd. But I am not sure and don't know how to fix it.

I cannot get any fixing method from google. It seems that this is not a popular issue.

Is there any one could share some good idea to solve it?

[SirRobin2318]

Is this the service file you're using?
(this one is a template from their github)

[bluephoenix]

the netatalk.service file I am using is as follows:

[SirRobin2318]

Hmmm, you're not alone apparently: https://groups.google.com/forum/#!topic/alt-f/G4P_o7Uikx0
Start by checking your /etc/hosts file, then you can try this:

[bluephoenix]

Hi, SirRobin,

Thanks a lot!

1. I cannot access the google forum since I am in China, where google is not legal. So I can not catch up your point about modify the file in "/etc/init.d" folder. Indeed, there is no the file with the same file name you mentioned in my folder.

2. If I start the netatalk by using the command "systemctl start netatalk" after booting in a terminal, CPU usage is normal. But if I let the systemd to start netatalk in the booting procedure, the CPU usage will reach 100%.

[SirRobin2318]

Could you post your /etc/hosts, /etc/hostname, and the afpd config file (on google they were saying it was /etc/init.d, on gentoo it might be different) please?


If you'd like to access google, you might be interested in this: https://www.torproject.org/projects/torbrowser.html.en
Just download it, un tar it, start it with ./start-tor-browser

[bluephoenix]

Many thanks, Robin,

I did solve this issue just by modifying the "/etc/hosts" file. I ever changed this file for some reason before and I deleted "localhost" item. Now I try to add this item back to the file and restart my computer, everything is fine now!

So really many thanks!

Although google is not legal in China, I can still have a little access to the page you proposed finally. After long waiting for the web page transfer, I get some clue from that page. So I find the "localhost" item missing issue.

Solved, quite nice!

[SirRobin2318]

Ok great