Red Hat Bugzilla – Bug 444600
clvmd init can hang for tens of minutes on startup
Last modified: 2010-01-11 23:07:17 EST
Description of problem:
I saw this on a gulm cluster ater I rebooted all 4 nodes. Only one of the nodes
came back up right away. Two came back up after about 10 minutes and the fourth
is still stuck. Here's where:
Starting ccsd: [ OK ]
[WARNING]lustered mirror log:[WARNING]
Starting lock_gulmd:[ OK ]
Starting fence domain:[WARNING]
Starting clvmd: [ OK ]
I'll attempt to gather more info.
Version-Release number of selected component (if applicable):
Each time, it seem that one of the 4 nodes hangs indefinately, where the other
three do eventually come up. I'll attempt to get a stack trace.
+ initlog -q -n /etc/rc3.d/S24clvmd -s 'clvmd startup' -e 1
+ '[' serial '!=' verbose -a -z '' ']'
+ '[' serial = color ']'
+ echo -n '[ '
[ + '[' serial = color ']'
+ echo -n OK
OK+ '[' serial = color ']'
+ echo -n ' ]'
]+ echo -ne '\r'
+ return 0
+ return 0
+ '[' 0 -ne 0 ']'
Created attachment 304149 [details]
traces of the cluster processes during the hang
There are a suspiciously small number of clvmd threads in that dump. Is it
possible to see what gulm is doing ?
I'm guessing this bz is a dup of 447799.
*** This bug has been marked as a duplicate of 447799 ***