[Pacemaker] ccm crash: /var/run/heartbeat/ccm/ccm Permission denied

[Rainer Traut]

Andrew Beekhof schrieb:
> 
> On Jul 28, 2008, at 12:24 PM, Rainer Traut wrote:
> 
>> sys: Centos 5 x86_64, 2 nodes
>>
>> # rpm -qa|grep heartbeat
>> pacemaker-heartbeat-0.6.5-8.2
>> heartbeat-ldirectord-2.1.3-23.1
>> heartbeat-resources-2.1.3-23.1
>> heartbeat-common-2.1.3-23.1
>> heartbeat-2.1.3-23.1
>>
>> One cluster member constanly reboots with these logs:
>>
>> Jul 28 12:11:47 n02asp7 ccm: [8768]: ERROR: socket_wait_conn_new: 
>> unlink failure(/var/run/heartbeat/ccm/ccm): Permission denied
>> Jul 28 12:11:47 n02asp7 ccm: [8768]: ERROR: socket_wait_conn_new: 
>> trying to create in /var/run/heartbeat/ccm/ccm bind:: Permission denied
>> Jul 28 12:11:47 n02asp7 ccm: [8768]: ERROR: Can't create wait channel: 
>> Resource temporarily unavailable
>> Jul 28 12:11:47 n02asp7 heartbeat: [8756]: WARN: Managed 
>> /usr/lib64/heartbeat/ccm process 8768 exited with return code 1.
>> Jul 28 12:11:47 n02asp7 stonithd: [8771]: info: Signing in with 
>> heartbeat.
>> Jul 28 12:11:47 n02asp7 heartbeat: [8756]: EMERG: Rebooting system. 
>> Reason: /usr/lib64/heartbeat/ccm
>>
>> and:
>> # ls -la /var/run/heartbeat/ccm/ccm
>> srwxrwxrwx 1 hacluster haclient 0 12. Jul 14:04 
>> /var/run/heartbeat/ccm/ccm
> 
> what about the directories it in?
> what user is ccm running as?
> 

uid 90, gid 90:

Jul 28 12:11:47 n02asp7 heartbeat: [8756]: info: Starting child client 
"/usr/lib64/heartbeat/ccm" (90,90)
Jul 28 12:11:47 n02asp7 heartbeat: [8768]: info: Starting 
"/usr/lib64/heartbeat/ccm" as uid 90  gid 90 (pid 8768)

# ls -la /var/run/heartbeat/ccm
insgesamt 8
drwxr-x--- 2 root      root     4096 18. Jul 17:30 .
drwxr-xr-x 5 root      root     4096 28. Jul 12:13 ..
srwxrwxrwx 1 hacluster haclient    0 12. Jul 14:04 ccm

And indeed:
# chown hacluster:haclient /var/run/heartbeat/ccm/

fixes thes problem.

Let me guess, it's because ccm now drops root privileges?

Rainer