[Linux-HA] socket of lrmd

[Junko IKEDA]

Hi,

On Fri, 03 Jul 2009 10:59:21 +0900, Junko IKEDA <ikedaj at intellilink.co.jp> wrote:

> Hi Dejan,
>
> Your patch could stop the error message from LVM RA.
> Many thanks!
> But I run Heartbeat 2.1.4 so I worry about whether 2.1.4 still have a problem about stonithd that you pointed.
>
> By the way, does closefiles() close both of the connections for syslogd and ha_logd ?

The patch works well.
but, what will happen if we don't use it?
It seems that lrmd keeps the file descriptor (3) for ha_logd/syslogd while running.
Will the number of the file descriptor reach its max limit without this patch?

Thanks,
Junko

>
> On Thu, 02 Jul 2009 22:04:06 +0900, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
>
>> Hi,
>>
>> On Thu, Jul 02, 2009 at 06:40:26PM +0900, Junko IKEDA wrote:
>>> Hi,
>>>
>>> I try to strace of lrmd.
>>>
>>> (1) running syslog, and strace
>>> # service syslog start
>>> # strace -o /tmp/strace-syslog-start.txt /usr/lib64/heartbeat/lrmd -rv
>>>
>>> socket(PF_FILE, SOCK_DGRAM, 0)          = 3
>>> fcntl(3, F_SETFD, FD_CLOEXEC)           = 0
>>> connect(3, {sa_family=AF_FILE, path="/dev/log"...}, 110) = 0
>>> sendto(3, "<30>Jul  2 17:02:46 lrmd: [16211"..., 102, MSG_NOSIGNAL, NULL, 0) = 102
>>>
>>> (2) stop syslog and strace
>>> # service syslog stop
>>> # strace -o /tmp/strace-syslog-stop.txt /usr/lib64/heartbeat/lrmd -rv
>>>
>>> socket(PF_FILE, SOCK_DGRAM, 0)          = 3
>>> fcntl(3, F_SETFD, FD_CLOEXEC)           = 0
>>> connect(3, {sa_family=AF_FILE, path="/dev/log"...}, 110) = -1 ENOENT (No such file or directory)
>>> close(3)                                = 0
>>> open("/dev/console", O_WRONLY|O_NOCTTY) = 3
>>>
>>> (for more details, see the attached files)
>>>
>>> Is the socket of lrmd for syslog?
>>> wild guess?
>>
>> Right. See lib/lrm/racommon.c:closefiles(). I must say I can't
>> remember why exactly was the connection to syslogd/ha_logd
>> needed. There has been a problem with stonithd, but now this code
>> is not in that execution path anymore. Perhaps you can test this
>> patch:
>>
>> diff -r 731f8f7b5450 lib/lrm/racommon.c
>> --- a/lib/lrm/racommon.c        Tue Jun 30 12:02:16 2009 +0200
>> +++ b/lib/lrm/racommon.c        Thu Jul 02 15:02:55 2009 +0200
>> @@ -145,7 +145,7 @@
>>
>>         /* close all descriptors except stdin/out/err and
>>  * channels to logd */
>>         for (fd = getdtablesize(); fd > STDERR_FILENO; fd--) {
>> -               if (!cl_log_is_logd_fd(fd))
>> +               /*if (!cl_log_is_logd_fd(fd))*/
>>                         close(fd);
>>         }
>>  }
>>
>> Thanks,
>>
>> Dejan