Cross Cluster Database Replication - adding GA docs#3107
Conversation
c9e0f8e to
cd79d46
Compare
|
Hi @NataliaIvakina, I've added a new commit: 8391014 So we've added the The reason for this change is to make it more distinct. With that said, should I remove the changes to the |
Hi @yirensum! |
|
@yirensum, sorry, my bad. Let's keep the syntax table as it is in the |
no problem, I've reverted that file to the dev version |
NataliaIvakina
added
DO NOT MERGE
2026.08
and removed
NOT_READY_FOR_MERGE
DO NOT MERGE
2026.07
labels
73ea59f to
ea9afa5
Compare
| Verify that the `foo-replica` database is online on the desired number of servers, in the desired roles. | ||
| Note that the type is labelled as `replica`. | ||
|
|
||
| If a `seedURI` is also specified, the initial seeding of the replica database is done from the `seedURI`. |
There was a problem hiding this comment.
We don't mention seedURI anywhere before. Why should we use it? And if we can use it, when and how to do it?
There was a problem hiding this comment.
so seedURI is one of the options of creating a database (see https://neo4j.com/docs/operations-manual/current/database-administration/standard-databases/seed-from-uri/), but maybe we should link it or create a new section here?
We can use it if we want to start the seeding of a database from a backup location usually.
There was a problem hiding this comment.
good, so there is also a way to create a replica from a database backup using seedURI. If you specify remote addresses then the replica pulls transactions from the upstream db using servers listed in addresses. If you specify pullURI, then the replica pulls diffs from there after initialization.
Am I right?
This way of creating a replica could be presented as a separate section. Let me think how to do it better.
There was a problem hiding this comment.
yes that's exactly so. Just to give further context (though we probably do not need to tell the customer this in the docs), internally we are applying transactions - either pulling transactions from addresses, or unpacking backups and getting transactions that way. As such ur mental model is correct.
| The differential backup chain must remain unbroken from the replica database's current point onward. | ||
|
|
||
| The user is responsible for continuing to create and push differential backups for the upstream database to the `pullURI` location. | ||
| If the differential backup chain is broken, for example because a required differential backup is missing, the replica database must be recreated from a complete backup chain. |
There was a problem hiding this comment.
This sentence is not super useful or clear: If the diff backup chain is broken ... the replica db must be recreated from a complete backup chain.
The second part of the sentence doesn't say what to do if the diff chain is broken.
But previously we said that a complete backup chain is a chain of a full backup + diffs (optionally). So, does this the replica db must be recreated from a complete backup chain mean that the replica must be recreated from the latest full backup?
At the same time, we state that a replica consumes only diffs after initialization.
There was a problem hiding this comment.
I see that this is not so clear, let me see if I can add something to the product to make it clearer about gaps in the chain. Will get back to you about this
| `remote`:: The name of the upstream database in a remote cluster. | ||
| `addresses`:: A list of xref:configuration/configuration-settings.adoc#config_dbms.cluster.endpoints[cluster endpoints] for the servers in the remote cluster. | ||
| Use this if there is a direct network connection between clusters. + | ||
| There is no requirement that every remote server hosts the upstream database. |
There was a problem hiding this comment.
Recently we added "There is no requirement that every remote server hosts the upstream database" to the option remoteAddresses of the internal.dbms.createReplicaDatabase() procedure. Should we use this clarification here?
There was a problem hiding this comment.
isn't this added to the create replica DDL rather than the internal.dbms.createReplicaDatabase?
# Conflicts: # modules/ROOT/pages/clustering/replicating-databases-across-clusters.adoc
Co-authored-by: Natalia Ivakina <82437520+NataliaIvakina@users.noreply.github.com>
ea9afa5 to
aad0b50
Compare
Co-authored-by: Natalia Ivakina <82437520+NataliaIvakina@users.noreply.github.com>
|
This PR includes documentation updates Updated pages: |
4 participants
In the PR, we add the GA docs of this feature.
We introduce:
dbms.promoteReplicaDatabaseNote that some images here have been ai-generated - they give a sense of how the image should look like but are placeholders and need to be replaced with proper images.
Part of CLUSTER-1587