Re: [Apcupsd-users] apcupsd 3.14.0 ether/net loses sync?

[Adam K. <akr...@ro...>]

Jan Ceuleers wrote:
> Adam,
>
> It's happened again.
>
> Adam Kropelin wrote:
>>> Would it help if, next time I notice the problem and before
>>> restarting the daemon, I try and attach an strace to the apcdev
>>> process, like so:
>>
>> Yes, that would be a good source of information. I'm very curious
>> where it is getting stuck.
>
> [root@skr03 root]# ps -aux --forest | grep apc
> root      1016  0.0  0.7 10224  964 ?        S    Mar21   0:05
> /sbin/apcupsd -f
> root      1027  0.0  0.7 10224  964 ?        S    Mar21   0:00  \_
> /sbin/apcupsd
> root      1032  0.0  0.7 10224  964 ?        S    Mar21   0:00      \_
> /sbin/apc

It is interesting that you appear to be running a different threading 
library than I am. On my newly-installed CentOS 3 vm I see only a single 
apcupsd process whereas you seem to show a process per thread, which is 
(or at least was; I've lost track) the typical linux threading model. I 
never did pay much attention to LinuxThreads vs. NPTL and such. I should 
research this a bit.

> [root@skr03 root]# strace -p 1027
> getppid()                               = 1016
> poll([{fd=4, events=POLLIN}], 1, 2000)  = 0
> getppid()                               = 1016
> poll([{fd=4, events=POLLIN}], 1, 2000)  = 0
> getppid()                               = 1016
> poll([{fd=4, events=POLLIN}], 1, 2000)  = 0
> getppid()                               = 1016
> poll([{fd=4, events=POLLIN}], 1, 2000)  = 0
> getppid()                               = 1016
> poll([{fd=4, events=POLLIN}], 1, 2000)  = 0
> getppid()                               = 1016
> poll( <unfinished ...>
> [root@skr03 root]#

That looks like the nis server thread you're stracing (fd #4 is the 
socket on which apcupsd is listening for nis connections, which is 
unused in your scenario). Could you strace pid 1032 instead?

> Checking open files:
>
> [root@skr03 root]# lsof | grep 3551
> apcupsd   1016 root    6u  IPv4       2691               TCP *:3551
> (LISTEN) apcupsd   1016 root    8u  IPv4      11953               TCP
> skr03.xperim.be:3512->penta.xperim.be:3551 (ESTABLISHED)
> apcupsd   1027 root    6u  IPv4       2691               TCP *:3551
> (LISTEN) apcupsd   1027 root    8u  IPv4      11953               TCP
> skr03.xperim.be:3512->penta.xperim.be:3551 (ESTABLISHED)
> apcupsd   1032 root    6u  IPv4       2691               TCP *:3551
> (LISTEN) apcupsd   1032 root    8u  IPv4      11953               TCP
> skr03.xperim.be:3512->penta.xperim.be:3551 (ESTABLISHED)

' lsof | grep apc' should show that fd #4 is a TCP socket on port 3551 
open for listening, which is the server side of nis and unfortunately 
uninteresting for this particular problem case.

--Adam