Confluence Data Center not compatible with MySQL Cluster

[Foucault Dupleix]

Hello,

I refer to this "feature" :

https://jira.atlassian.com/browse/CONFSERVER-20277)

We started a projet to setup a new high available and scalable plateform for both Jira and Confluence with Data Center evaluation licences.

As our current plateforms were using MySQL, we, of course, choosed MySQL Cluster as database system in order to ease the migration.

Unfortunately, when we tried to install the Confluence application, we got the "Could not read a HI value, you need to populate the table: "hibernate_unique_key"..." error. And manual actions do not solve the issue.

Can you please explain how you can propose Data Center licences for high availability if the database becomes a Single Point Of Failure ?

Will you update this page [https://confluence.atlassian.com/doc/supported-platforms-207488198.html] to explain that you do not support clustered database ?

Did some of your clients identify a workaround ?

I saw that some users succeeded with an Oracle RAC + Data Guard architecture. Do you officially support this setup ? 

Oracle licences are more expensive than MySQL, so the cost of the projet would increase a lot. Does anybody have another alternative to propose? 

Thanks,

Regards,

Foucault DUPLEIX

[Danny van den Berg]

Thank you for your reply.

I will try some workaround and otherwise just switch over to Oracle RAC maybe. We already have setup Confluence with PostgreSQL and PGPool-II, but we don't have any floating ip-adres available so still no HA on database level..

Oracle than will the solution, but hardening PostgreSQL or MySQL and make sure it won't fail still has then my preference..

[Bert Kelder]

Thanks for sharing

[Danny van den Berg]

Same issue here. Created a MySQL INNODB Cluster to achieve HA but setting up Confluence with a cluster is not working at all..

Please Atlassian, also make sure that the DB will become high available.....

[Foucault Dupleix]

Thank you Dany for your sharing.

To be more accurate for those who face the same situation :

We first tried a MySQL Groupe Replication technology. But it requires that all tables have a primary key. One of the Confluence table does not have a primary key (it's a shame by the way...), so the installation failed.

Then, we tried a MySQL replication and the installation went well. The High Availability is not as good as in Group Replication technology in case of Master problem or data corruption, but it's not so bad neither. 

The Data Center licences are very too expensive compared to this possibility of single point of failure... 

[Bert Kelder]

Can you share a link abt users succeeding with Oracle RAC/Dataguard. We've been discussing HA DB setup with Atlassian on last Summit, but they do not have any plans on this. We're actually looking for other people / companies using Confluence DC with a HA DB setup. We also looked at Postgres with PGPoolII but PGPool is not stable enough for this (yet)

[Foucault Dupleix]

In the following "Consideration", in the comments, you can find a testimony from David Hergert saying tht he used Oracle RAC + Dataguard (it's from 2017 and he does not say if it works without issues) :

https://jira.atlassian.com/browse/CONFSERVER-20277)

[Aleksei Faians]

There are actually 3 tables without primary keys, at least in version 7.8.0. I didn't try with more recent ones. Why Atlassian cannot add a primary key into three tables for almost three years is a big question.

mysql> select tab.table_schema as database_name,
-> tab.table_name
-> from information_schema.tables tab
-> left join information_schema.table_constraints tco
-> on tab.table_schema = tco.table_schema
-> and tab.table_name = tco.table_name
-> and tco.constraint_type = 'PRIMARY KEY'
-> where tco.constraint_type is null
-> and tab.table_schema = 'confluence_dev'
-> order by tab.table_schema,
-> tab.table_name;
+----------------+----------------------+
| database_name | table_name |
+----------------+----------------------+
| confluence_dev | hibernate_unique_key |
| confluence_dev | MIG_DB_CHANGELOG |
| confluence_dev | storageenginetest |
+----------------+----------------------+
3 rows in set (0.48 sec)

[Aleksei Faians]

Here's another request specifically about InnoDB cluster incompatibility: https://jira.atlassian.com/browse/CONFSERVER-68724