Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 05/21/2012 10:39 AM, 陈楠 wrote: > I configure it like I show. I create metadata on every node and set one > disk state to optodate and the other inconsistent, then drbd can sync > and heartbeat can failover. Can you tell me the reason? I need the > reason to persuade my leader. Dan has shortly described the reason, as have I in my first mail to this thread: > The idea with the DRBD configs is to tell each node about the whole > picture, and each node will pick the information it needs. That's why it > can even work when each node sees a slightly different picture. > But don't try tricking DRBD, because you will only confuse yourself. > Just install the whole, working config on each of your nodes. Make sure to explain to your manager that installing different configs on each DRBD peer is a maintenance nightmare. E.g., in a year, you will probably not be able to tell what's supposedly correct anymore. There is really no reason to set yourself up for future silly mistakes. Is there any reason your management insists on a messed up configuration?