# Moving to from DRBD 7 to DRBD 8

**URL:** <https://forums.linbit.com/t/moving-to-from-drbd-7-to-drbd-8/1293>\
**Category:** General\
**Tags:** drbd\
**Created:** [October 13, 2026, 3:01pm UTC](https://forums.linbit.com/t/moving-to-from-drbd-7-to-drbd-8/1293 "2026-10-13T15:01:33Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![YogliMogli](https://avatars.discourse-cdn.com/v4/letter/y/f0a364/32.png) [@YogliMogli](https://forums.linbit.com/u/YogliMogli)\
**Post date:** [October 13, 2026, 3:01pm UTC](https://forums.linbit.com/t/moving-to-from-drbd-7-to-drbd-8/1293/1 "2026-10-13T15:01:33Z")

</div>

So in HA + DR (Three DBs: one HA pair [Primary \<\> Secondary] and one DR [i.e. DR Secondary]) there is one ‘SOURCE-of-TRUTH’ which is the Primary DB. When you see UpToDate in a Secondary (drbdadm status all) you know that the UPToDate means that the Secondary has an EXACT-, SIMILAR COPY to the Souce-of-Truth (Primary) and this connection made via tcp/7788.

HOWERVER, I also see the TWO Secondaries (HA Secondary \<\> DR Secondary) communicate with each other (via port 7789) and here are my question:

1. Why do they communicate each other ? What are the ‘benefits’ of two Secondaries communcate with each other if the source-of-truth is the Primary ?

2. Are they ‘watching’ each other ? For what ? I know there is the ‘mesh structure’ between all nodes but what benefits it gives us if everything is UpToDate relative to the Primary ? Are they need to be also ‘UpToDate’ compare to the copy of the other Secondary (i.e. two Secondaries compare the copies they have).

3. Finally, what kind of data they resync with each other via port tcp/7789 (bitmap and other Meta data) or somthing ‘else’ ?

Again, I will be happy for any link / clip to explain this idea.

Thanks.

---

<div class="post-metadata">

**Author:** ![Devin](https://yyz1.discourse-cdn.com/flex003/user_avatar/forums.linbit.com/devin/32/26_2.png) [@Devin](https://forums.linbit.com/u/Devin)\
**Post date:** [October 14, 2026, 10:54pm UTC](https://forums.linbit.com/t/moving-to-from-drbd-7-to-drbd-8/1293/2 "2026-10-14T22:54:13Z")

</div>

The primary reason DRBD v9 works on a full-mesh is because the Primary node can move, and indeed it should if one of the nodes becomes unreachable or behaves unexpectedly.

If not a full-mesh then what?. The primary is now dead, and one of the secondaries should become primary, but it can no longer reach the other peer, because we’re not full mesh, and I needed that peer to complete the connection chain, but it is now gone.

This is a very simplified example, and there is more to it then that.

1. What do the two primaries do when the single “source-of-truth” is gone? If one was to go primary when disconnected from the peer you would have a split-brain. All nodes are connected to each other to handle circumstances which would be considered outside of “normal operation”.

2. Consider a situation where we use quorum to prevent split-brains. Consider a situation when you have three secondaries connected (for example after all three nodes are freshly rebooted). We need mechanisms where all nodes can compare with/against each other. Again, you speak about everything relative to the Primary, but what when there is no Primary?

3. It depends on the situation. Under normal operation it’s just some keep-alive type pings. It’s when the primary is missing or a node join/leaves the cluster that more information is compared.

I hope this help explains it, but if there is anything specifically you want more detailed on let me know and I will see what I can dig up.