Locking Issues

[Markus Iturriaga]

Hi all,

I know locking issues have come up before on this list, but I'd like to  
take another stab and finding out what might be wrong with out cfengine  
(2.1.14) setup. We seem to be getting a number of random "Another  
cfengine seems to have done XYZ" errors as well as a few consistent  
ones.

However, if we check the run log, it claims the locks had been removed.  
The locking error (on copy) we get consistently is:

Checking copy from  
localhost:/var/cfengine/inputs/configs/linux/debian_based/debian/3_1/ 
etc to /etc
Checking copy from  
localhost:/var/cfengine/inputs/configs/linux/debian_based/debian/etc to  
/etc
cfengine:lemon: Another cfengine seems to have done  
[copy._var_cfengine_inputs_configs_linux_debian_based_de__etc_localhost] 
 since I started (elapsed=-1)

Checking the run log we see:

Thu Jun  9 12:03:22 2005:Lock removed normally
:pid=30443: 
lock.cfagent_conf.lemon.copy._var_cfengine_inputs_configs_linux_debian_b 
ased_de__etc_localhost_188:
Thu Jun  9 12:03:22 2005:Lock removed normally
:pid=30443: 
lock.cfagent_conf.lemon.copy._var_cfengine_inputs_configs_linux_debian_b 
ased_et__etc_localhost_3604:
Thu Jun  9 12:03:22 2005:Lock removed normally
:pid=30443: 
lock.cfagent_conf.lemon.copy._var_cfengine_inputs_configs_linux_etc__etc 
_localhost_2020:
Thu Jun  9 12:03:22 2005:Lock removed normally
:pid=30443: 
lock.cfagent_conf.lemon.copy._var_cfengine_inputs_configs_etc__etc_local 
host_3202:

We also ran cfagent with -d 5 and looked at the output and for this  
lock,
it seems to just be acting normally.  We have gone through our configs
and looked to see if this could be defined to happen twice, but
haven't been able to find anything yet.  Also, if it were, it seems like
it would break for the other directories getting copied as well, since
they are all defined together at once.

I wondered if we were running into a string-length issue, but it  
appears other directories with even longer paths do not have problems.  
Running cfagent without locks seems to work fine, but that seems risky  
to me.

Any suggestions would be greatly appreciated.

---
Markus A. Iturriaga Woelfel, SysAdmin
Department of Computer Science
University of Tennessee, Knoxville
miturria@cs.utk.edu / (865) 974-3837