[Linux-ha-dev] Overflow on longclock_t?

[Guillem Anguera]

Hi list!

I asked this on main mailing list, but nobody seems to know it...

I'm using heartbeat on 2 active/active firewall systems; within the last 
48 hours, coinciding with an uptime of 49 days and a few hours, all 
servers have suffered the same problem: /var/log/heartbeat.log grows 
until fills /var free space partition with messages like attached file. 
 From the first message, the other node take over all resources despite 
of the original node isn't able to release it, at this point original 
node doesn't works.

I watch at source code (include/clplumbing/longclock.h) that longclock_t 
is at least defined as 64 bits variable, that seems to be enough. But I 
think that on my servers is defined as 32 bits variable:

2^32 = 4294967296 / 1000 (miliseconds to seconds) = 4294967,296 / 3600 
(seconds to hours) = 1193,046471111 / 24 (hours to days) = 49,71026963 
days, like system's uptime.

What do you think? Is that possible?

Additional Information:
Debian version: Sarge (3.1)
Vanilla kernel version: 2.4.34.5
Debian heartbeat version: 2.0.7-2

P.D: Sorry for my poor english skills.

-- 
Guillem Anguera
Administrador de Sistemas
Jazztel - DATAGRAMA
Tel: 900 80 83 80
Fax: +34 93 289 63 10
Em at il: ganguera () datagrama ! net
http://www.jazztel.es

-------------- next part --------------
A non-text attachment was scrubbed...
Name: heartbeat.log
Type: text/x-log
Size: 4988 bytes
Desc: not available
Url : http://lists.community.tummy.com/pipermail/linux-ha-dev/attachments/20080328/5ed02b5e/heartbeat.bin