[DRBD-user] Ways to force a heartbeat take-over on io-error?

Dave Dykstra dwdha at drdykstra.us
Thu Jul 14 20:32:24 CEST 2005

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Thu, Jul 14, 2005 at 10:18:30AM +0200, Tim Bruijnzeels wrote:
> Hi,
> 
> 
> I am using:
> drbd 0.7.10
> heartbeat 1.2.3
> RedHat EL 4
> 
> 
> I am in the midst of setting up a fail-over cluster. The way I am 
> setting it up it's actually a two-way failover cluster. During normal 
> operation node A will host a specific set of services and be the drbd 
> primary for the associated drbd disk, and node B similar for another set.
> For the OS level partitions I use software raid. As drdbd effectively 
> mirrors the data between two systems I thought I'd save some disk space 
> and just use unmirrored partitions for the drbd devices on both nodes...
> 
> The problem is of course that when a single disk crashes I will have to 
> force a take-over.
> 
> As far as I can tell I can use the following option in the 'disk' 
> section to force this:
> 
>  on-io-error panic
> 
> The panic will force heartbeat on the other node to initiate a take over.
> 
> But is there another way? E.g. can I somehow make drbd invoke a 
> 'heartbeat stop' on-io-error; and preferably send out an alert message? 
> That way it would be easier for me to log into the crashed node and 
> investigate..
> 
> The more I think about it, the more I am starting to get in favor of the 
> 'panic' option to be on the safe side. But I thought I'd ask my question 
> here to see if anyone else has ideas / experience on/with this.
> 
> Cheers,
> 
> Tim


I hadn't noticed that before, but it sounds like a good idea.  I wonder
what happens if there's a hard error on the disk though: will the computer
keep rebooting as soon as it gets to the point of starting drbd?

- Dave



More information about the drbd-user mailing list