Hi, On Tue, Jun 10, 2008 at 12:46:09PM -0500, Matt Zagrabelny wrote: > On Tue, 2008-06-10 at 19:07 +0200, Dejan Muhamedagic wrote: > > Hi, > > > > On Tue, Jun 10, 2008 at 10:51:47AM -0500, Matt Zagrabelny wrote: > > > Greetings, > > > > > > I have a two node cluster (squash and turnip). > > > > > > +------------+ +------------+ > > > | /dev/ttyS1 | <====> | /dev/ttyS1 | > > > | | | | > > > | squash | | turnip | > > > | | | | > > > | eth2 | <====> | eth2 | > > > | (10.0.0.1) | | (10.0.0.2) | > > > +------------+ +------------+ > > > > > > squash% cat /etc/ha.d/ha.cf > > > use_logd on > > > > > > keepalive 1 > > > deadtime 5 > > > initdead 120 > > > > > > udpport 694 > > > baud 115200 > > > serial /dev/ttyS1 > > > bcast eth2 > > > > > > node squash > > > node turnip > > > > > > ping 131.212.4.166 > > > ping 192.168.115.46 > > > > > > crm on > > > > > > > > > It appears that the heartbeat module does not recover when I issue > > > 'ifdown eth2 && sleep 10 && ifup eth2'. > > > > Which version do you run? This should've been fixed in 2.1.3. See > > http://developerbugs.linux-foundation.org/show_bug.cgi?id=1732 > > for more details. > > % ssh squash dpkg -l heartbeat > ||/ Name Version Description > +++-==============-==============-============================================ > ii heartbeat 2.1.3-6 Subsystem for High-Availability Linux > > It does not look as though it is fixed (for me at least.) > > I downed and upped (squash# ifdown eth2 && sleep 10 && ifup eth2) the > interface 15 minutes ago and things still have not recovered - I still > see arp traffic for the broadcast address and cl_status reports the link > being down: > > squash% echo "squash eth2 - $(cl_status hblinkstatus squash eth2)" && > echo "turnip eth2 - $(cl_status hblinkstatus turnip eth2)" > squash eth2 - up > turnip eth2 - up > > turnip% echo "squash eth2 - $(cl_status hblinkstatus squash eth2)" && > echo "turnip eth2 - $(cl_status hblinkstatus turnip eth2)" > squash eth2 - dead > turnip eth2 - up > > I do not know what else I can do to debug this, aside from delving into > the source. > > Suggestions welcome.
According to the bugzilla, if the interface is gone the heartbeat write process should exit. Then the master control process should start a new one. Can you please reopen the bugzilla and attach a hb_report. Thanks, Dejan _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
