[Linux-HA] crmd would dumped core, while recovering from split-brain

[Andrew Beekhof]

On 9/3/07, Junko IKEDA <ikedaj at intellilink.co.jp> wrote:
> Hi,
>
> When some nodes recover from Split-Brain,
> crmd would dumped core.
> Does it work as required?
> Is crmd going to shutdown if membership instance ID went backward?
>
> heartbeat[2635]: 2007/09/03_11:48:21 info: Link prec370d:eth2 up.
> ...
> heartbeat[2635]: 2007/09/03_11:48:22 info: Status update for node prec370d:
> status active
> crmd[2648]: 2007/09/03_11:48:22 notice: crmd_ha_status_callback: Status
> update: Node prec370d now has status [active]
> ...
> crmd[2648]: 2007/09/03_11:48:23 info: crmd_ccm_msg_callback: Quorum
> (re)attained after event=NEW MEMBERSHIP (id=1)
> crmd[2648]: 2007/09/03_11:48:23 ERROR: crmd_ccm_msg_callback: Membership
> instance ID went backwards! 3->1
> cib[2644]: 2007/09/03_11:48:23 info: cib_ccm_msg_callback: PEER: prec370e
> crmd[2648]: 2007/09/03_11:48:23 ERROR: crm_abort: crmd_ccm_msg_callback:
> Triggered fatal assert at callbacks.c:526 : current_ccm_membership_id <=
> membership->m_instance

^^^^^^^^^

# man abort

the ccm is feeding us garbage so the only sane thing to do it get out
immediately

/me returns to his vacation