Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 08/16/2012 07:38 PM, Bruce Wolfe, M.S.W., CIO wrote: > Thanks, Digimer, and greetings again. > > The configuration is: > Failover of primary to secondary. > Two nodes: one primary, one secondary in, as you say, a RAID 1 > configuration. > > So, all three get used? In our current configuration, we are only using > DRBD and Heartbeat. To get automatic recovery after failure, yes, you need all three. DRBD simply replicates raw data. That's it, nothing more. Promoting a secondary node to primary requires external actions, be it by the user or via another program. Heartbeat is deprecated and has no future. Anyone using it should be making near-term plans to get off of it. So let's take the right off the table. Corosync is a stand alone tool that handles cluster membership and message passing. It doesn't care what other programs do or how they use it's message passing capabilities. It is merely a communications tool. Specifically; It decided who can send and receive messages amoungst a group of machines. In our case, we want this so that pacemaker can coordinate actions. Pacemaker is a cluster resource manager. That is, it reacts to changes in cluster membership and, based on defined policies, decides to stop, start, migrate or otherwise act on services. It doesn't care *how* machines in the cluster come and go, only that they do. So in your use case, you would setup DRBD to replicate data. Next, you would configure corosync to say "these two nodes are members of cluster X". Then you tell Pacemaker; "When both nodes are available, make node 1's DRBD the primary and make node 2 secondary. However, if not 1 dies, promote node 2 to primary. When node 1 returns, demote node 2 and promote noed 1." Make sense? -- Digimer Papers and Projects: https://alteeve.com