[DRBD-user] Automatic split brain recovery policies - after-sb-0pri - discard-older-primary

Dominik Klein dk at in-telegence.net
Thu Jan 8 07:47:32 CET 2009

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

When falling asleep in the train yesterday, I think I got the answer to
this question.

> In case of the following situation:
> nodes node1 and node2 are disconnected
> At t0, split brain
> At t1, node1 becomes primary
> At t2, node1 becomes secondary
> At t3, node2 becomes primary
> At t4, node2 becomes secondary

So node1 is "old" and node2 is "young".

> At t5, node1 becomes primary
> At t6, node1 becomes secondary

Now node2 is old and node1 is young.

> At t7, node2 becomes primary
> At t8, node2 becomes secondary

Now node1 is old again.

Keep on doing this and I guess you have your answer. Please correct me
if that's wrong and falling asleep did not solve this question :)


> At t9, node1 becomes primary
> At t10, node1 becomes secondary
> At t11, node2 becomes primary
> At t12, node2 becomes secondary
> At t13, node1 becomes primary
> At t14, node1 becomes secondary
> At t15, node2 becomes primary
> At t16, node2 becomes secondary
> At t17, node1 becomes primary
> At t18, node1 becomes secondary
> At t19, node2 becomes primary
> At t20, node2 becomes secondary
> At t21, node1 becomes primary
> At t22, node1 becomes secondary
> At t23, node2 becomes primary
> At t24, node2 becomes secondary
> At t25, split brain is fixed
> "discard-older-primary" policy will make node2 from t23 the sync source
> "discard-younger-primary" policy will make node1 from t1 the sync sourc,
> or node1 from another time than t1 ?

More information about the drbd-user mailing list