[Linux-HA] Heartbeat with serial port refuses to start

[Jordi Moreno]

Hi all,

I have searched for info about this issue with no success...

I have been using Heartbeat2 in various environments with no problems at 
all. Recently, I started to try version 3.0.3 of Heartbeat. We need some 
special requirements, like having the configuration files in a special 
directory, so I compiled Cluster Glue, Resource Agents, Heartbeat and 
Pacemaker with the option --sysconfdir. Anyway, I think it doesn't have 
anything to do with this issue, because I get the same error compiling 
without this option.

When I try to start Heartbeat, it fails saying it couldn't lock the 
serial line:

...
Oct 18 18:35:09 SCBR0097 heartbeat: [20625]: info: Configuration 
validated. Starting heartbeat 3.0.2
Oct 18 18:35:09 SCBR0097 heartbeat: [20625]: debug: HA configuration OK. 
  Heartbeat starting.
Oct 18 18:35:09 SCBR0097 heartbeat: [20625]: info: Heartbeat Hg Version: 
node: 645cec2ec68eb0cd41aa12ce282a23df45885561
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: info: heartbeat: version 3.0.2
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: info: Heartbeat generation: 
1285920283
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: debug: uuid 
is:48bc3ad5-1fdf-42ea-8c7b-1408b14232af
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: debug: FIFO process pid: 20629
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: debug: 
G_main_IPC_Channel_constructor(sock=3,3)
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: debug: opening serial 
/dev/ttyS0 (serial ring)
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: ERROR: glib: cannot lock 
line /dev/ttyS0
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: ERROR: make_io_childpair: 
cannot open serial /dev/ttyS0
Oct 18 18:35:09 SCBR0097 heartbeat: [20626]: debug: Exiting from pid 
20626 [rc=1]
Oct 18 18:35:10 SCBR0097 heartbeat: [20629]: debug: pid 20629 locked in 
memory.
Oct 18 18:35:10 SCBR0097 heartbeat: [20629]: debug: Limiting CPU: 6 CPU 
seconds every 59999 milliseconds
Oct 18 18:35:11 SCBR0097 heartbeat: [20629]: CRIT: Emergency Shutdown: 
Master Control process died.
Oct 18 18:35:11 SCBR0097 heartbeat: [20629]: CRIT: Killing pid 20626 
with SIGTERM
Oct 18 18:35:11 SCBR0097 heartbeat: [20629]: CRIT: Emergency 
Shutdown(MCP dead): Killing ourselves.
Oct 18 18:35:11 SCBR0097 heartbeat: [20629]: debug: Process 20629 
processing SIGTERM
Oct 18 18:35:11 SCBR0097 heartbeat: [20629]: debug: Exiting from pid 
20629 [rc=15]

If I try to start Heartbeat without the serial option, and it runs with 
no problems at all. I have checked permissions, and eventually changed 
/dev/ttyS0 to 777 (dangerous, I know, but just to be sure). Serial 
communication runs OK between the machines (tested cat in one machine, 
then echo in the other, and it works). I would like to know why 
Heartbeat can't lock the line, but even in debug level 10 it doesn't 
tell the cause. I have tried strace too, but the only relevant line 
seems to be the following:

stat64("/dev/ttyS0", {st_mode=S_IFCHR|0660, st_rdev=makedev(4, 64), 
...}) = 0

Maybe Heartbeat calls another program that fails, I don't know...

Sorry for the long post but I have googled for a long time and I'm a bit 
desperate... thanks in advance!

---
Un saludo,

       Jordi Moreno
       CIM Internet, S.L.
       jmoreno at cim.es