[Linux-ha-dev] ipfail and ping
gshi at ncsa.uiuc.edu
Sat Oct 16 16:24:20 MDT 2004
At 09:36 PM 10/15/2004 -0600, you wrote:
>Guochun Shi wrote:
>>At 12:00 AM 10/15/2004 -0400, you wrote:
>>>On Thu, Oct 14, 2004 at 03:47:50PM -0500, Guochun Shi wrote:
>>>>Since ipfail has to work with ping/pinggroup/hbaping or makes no sense,
>>>>maybe we should give error messages and quit if ipfail is configured without ping/pinggroup/hbaping in ha.cf
>>>It could be done. It doesn't break anything for someone to leave out
>>>ping/pinggroup/hbaping directives. ipfail will behave exactly as
>>Yes, it does not hurt if a user does not configure any ping node, but obviously that would be a misconfiguration
>>because it does not do what ipfail is designed to do. It would be better if we can prevent that in
>>>Unfortunately it's not as easy as checking num_ping_nodes > 0, since
>>>it's fine for there to be 0 (when they're dead.) I'm thinking it might
>>>be annoying to keep track of the various kinds of ping nodes and whether
>>>they've been configured in ha.cf.
>>I think it's the configuration we care. It does not matter whether ping nodes are dead or not. As long as we see ipfail is configured and no ping nodes are in ha.cf, we print out error messages.
>>The modifications should be in config.c if this is to be done.
>I disagree with this last statement.
>Heartbeat has no knowledge of the requirements of any programs that it might start up - and it has no business knowing what they are.
But we know it :). It would be nice if we can add the modification and prevent a user from mis-configuring ipfail.
If you think it's not heartbeat's business to do this extra checking, then it's fine with me
More information about the Linux-HA-Dev