On 07/30/2010 02:09 PM, Ofer Inbar wrote:
> I forgot to add something useful, appended below.
> -- Cos
>
> On Fri, Jul 30, 2010 at 05:02:04PM -0400, I wrote:
>> I've gotten further, into another roadblock:
>> Starting the cman service gets as far as successfully running ccsd,
>> but then at the next step where it checks cman status, it fails.
>>
>> | Starting cluster:
>> | Loading modules... done
>> | Mounting configfs... done
>> | Starting ccsd... done
>> | Starting cman... failed
>> |
>> | [FAILED]
>>
Is aisexec already running (ie: did you set it to start via init
script?). cman starts aisexec, and therefore, the initscript for
aisexec should be disabled.
system logs would be helpful.
Regards
-steve
>> Any attempt to use cman_tool gives:
>> cman_tool: Cannot open connection to cman, is it running ?
>>
>> strace shows that it fails here:
>> connect(3, {sa_family=AF_FILE, path="/var/run/cman_client"}, 110) = -1
>> ECONNREFUSED (Connection refused)
>>
>> strace -fp on ccsd shows the same thing, in a loop, like this:
>>
>> [pid 7670] rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0
>> [pid 7670] rt_sigaction(SIGCHLD, NULL, {SIG_DFL, [], 0}, 8) = 0
>> [pid 7670] rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
>> [pid 7670] nanosleep({1, 0}, {1, 0}) = 0
>> [pid 7670] socket(PF_FILE, SOCK_STREAM, 0) = 9
>> [pid 7670] fcntl(9, F_SETFD, FD_CLOEXEC) = 0
>> [pid 7670] connect(9, {sa_family=AF_FILE, path="/var/run/cman_client"},
>> 110) = -1 ECONNREFUSED (Connection refused)
>> [pid 7670] close(9) = 0
>>
>> Named sockets are there and seem to have the right permissions:
>> srw------- 1 root root 0 Jul 30 15:21 /var/run/cman_admin=
>> srw-rw---- 1 root root 0 Jul 30 15:21 /var/run/cman_client=
>>
>> I've been googling for ideas. I don't really understand what's
>> supposed to using this named socket to communicate with what,
>> but I think this is a way for clients to talk to the cman plugin
>> in aisexec. aisexec seems okay on all my cluster nodes, and they're
>> all communicating with each other, but I also don't know the right
>> ways to poke at it to check its health.
>>
>> I'm continuing to experiment, and Google for ideas, but maybe
>> someone here can point me in the right direction.
>> -- Cos
>
> I noticed that lsof on my aisexec daemons on this cluster shows that
> they do *not* have /var/run/cman_client or /var/run/cman_admin open.
> lsof shows nothing has those open. When I checked on a working
> cluster I saw that aisexec does have both of those open.
>
> I tried rm'ing both and letting them be recreated when I restarted the
> aisexec and cman services, but I still get ECONNREFUSED on cman_client
> when I try running any cman_tool command.
> -- Cos
> _______________________________________________
> Openais mailing list
> [email protected]
> https://lists.linux-foundation.org/mailman/listinfo/openais
_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais
