Hi Dejan, Actually I don't need CCM ,I tried to write my private CCM just leverage Heartbeat's messaging channel and heartbeat function. I want a my private memberhsip management. Is it possible? I think heartbeat layer supports multiple nodes which means heartbeat can sendclustermsg() to multiple nodes as long as heartbeat add the node into the node list and detect node status for client by periodically checking status of the nodes.
My understanding is heartbeat should support multiple nodes even without CCM. But I need implement my own cluster membership and resource management. Am I correct? I did a hack to set DoManageResources to FALSE, it seems works. I try to figure out a way no need to hack code inside heartbeat. Could you help me? Thanks Javen 2010/1/5 Dejan Muhamedagic <[email protected]> > Hi, > > On Tue, Jan 05, 2010 at 07:15:24PM +0800, Javen Wu wrote: > > The system was reboot automatically after I set crm to "yes" :( > > Use "crm respawn" instead of "crm yes" then. > > > ====%<=== > > Jan 05 19:08:52 na40-58 crmd: [6104]: info: crmd_init: Starting crmd > > Jan 05 19:08:52 na40-58 crmd: [6104]: info: G_main_add_SignalHandler: > Added > > signal handler for signal 17 > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: WARN: Managed > > /usr/lib64/heartbeat/attrd process 6103 killed by signal 11 [SIGSEGV - > > Segmentation violation]. > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: ERROR: Managed > > /usr/lib64/heartbeat/attrd process 6103 dumped core > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: ERROR: Client > > /usr/lib64/heartbeat/attrd (pid=6103) killed by signal 11. > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: ERROR: Respawning client > > "/usr/lib64/heartbeat/attrd": > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: info: Starting child client > > "/usr/lib64/heartbeat/attrd" (496,119) > > Jan 05 19:08:52 na40-58 cib: [6100]: info: retrieveCib: Reading cluster > > configuration from: /usr/var/lib/heartbeat/crm/cib.xml (digest: > > /usr/var/lib/heartbeat/crm/cib.xml.sig) > > Jan 05 19:08:52 na40-58 cib: [6100]: info: startCib: CIB Initialization > > completed successfully > > Jan 05 19:08:52 na40-58 cib: [6100]: info: crm_cluster_connect: > Connecting > > to Heartbeat > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: WARN: Managed > > /usr/lib64/heartbeat/cib process 6100 killed by signal 11 [SIGSEGV - > > Segmentation violation]. > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: ERROR: Managed > > /usr/lib64/heartbeat/cib process 6100 dumped core > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: ERROR: Client > > /usr/lib64/heartbeat/cib (pid=6100) killed by signal 11. > > Jan 05 19:08:52 na40-58 heartbeat: [6091]: EMERG: Rebooting system. > Reason: > > /usr/lib64/heartbeat/cib > > > > ====%<===== > > > > Is there a bug? > > Most probably. There both cib and attrd segfault. There should be > coredumps. Don't know which version of Pacemaker do you run, > perhaps those were already fixed. > > > Is it possble I disable CCM and pacemaker software stack and only run my > > client build on top of heartbeat service? > > No. In particular, without CCM you can't have more than two > nodes. It is probably possible to run heartbeat with CCM but > without CRM (Pacemaker), but then you'll have to do some changes > in the heartbeat code to enable that. > > Thanks, > > Dejan > > > Thanks > > Javen > > > > > > > > 2010/1/5 Dejan Muhamedagic <[email protected]> > > > > > Hi, > > > > > > On Tue, Jan 05, 2010 at 06:50:38PM +0800, Javen Wu wrote: > > > > Hi Folks, > > > > > > > > I wrote a simple client by use of the Heartbeat client API and build > it > > > on > > > > the top of the heartbeat 3.0. > > > > My client is to form a simple private multi-node cluster and support > > > > autojoin. > > > > > > > > ------%<--segment of my ha.cf------ > > > > autojoin any > > > > > > > > apiauth myclient uid=root > > > > respawn root /usr/lib64/heartbeat/myclient > > > > ------%<------------------------------------- > > > > > > > > But I found the below error when I run my client. > > > > =========%<============ > > > > heartbeat[21726]: 2010/01/05_05:51:54 WARN: Deprecated 'legacy' > > > > auto_failback option selected. > > > > heartbeat[21726]: 2010/01/05_05:51:54 WARN: Please convert to > > > 'auto_failback > > > > on'. > > > > heartbeat[21726]: 2010/01/05_05:51:54 WARN: See documentation for > > > conversion > > > > details. > > > > heartbeat[21726]: 2010/01/05_05:51:54 WARN: logd is enabled but > > > > logfile/debugfile/logfacility is still configured in ha.cf > > > > heartbeat[21726]: 2010/01/05_05:51:54 ERROR: R1 style resource > management > > > > conflicts with autojoin set > > > > heartbeat[21726]: 2010/01/05_05:51:54 ERROR: You need either unset > > > autojoin > > > > or enable crm > > > > heartbeat[21726]: 2010/01/05_05:51:54 ERROR: Configuration error, > > > heartbeat > > > > not started. > > > > heartbeat[21726]: 2010/01/05_05:51:54 debug: Exiting from pid 21726 > > > [rc=6] > > > > ======%<=============== > > > > > > > > After check the code, I found the failure caused by > > > > heartbeat/config.c:r1_style_valid(). > > > > I don't know why my client was thought as "r1" (release 1) style by > > > default > > > > and the "autojoin any" must be disabled in r1 style. > > > > > > > > Do you know how to make a "non-r1" style client and make my client > > > support > > > > autojoin and multiple node at same time? > > > > > > Add "crm yes" to ha.cf. > > > > > > Thanks, > > > > > > Dejan > > > > > > > Thanks > > > > -- > > > > Javen Wu > > > > _______________________________________________ > > > > Linux-HA mailing list > > > > [email protected] > > > > http://lists.linux-ha.org/mailman/listinfo/linux-ha > > > > See also: http://linux-ha.org/ReportingProblems > > > _______________________________________________ > > > Linux-HA mailing list > > > [email protected] > > > http://lists.linux-ha.org/mailman/listinfo/linux-ha > > > See also: http://linux-ha.org/ReportingProblems > > > > > > > > > > > -- > > Javen Wu > > _______________________________________________ > > Linux-HA mailing list > > [email protected] > > http://lists.linux-ha.org/mailman/listinfo/linux-ha > > See also: http://linux-ha.org/ReportingProblems > _______________________________________________ > Linux-HA mailing list > [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems > -- Javen Wu _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
