Master Region and Replication


#1

I have been playing around with regions and having a master region to see all regions in Darga. I have one region that is located farther away than others, and is having issues replicating. In looking at the logs it appears the connection resets on the database before the replication completes. I can see that the replication is active, but the backlog continues to increment. Is there a setting that I can mess with to keep the ruby replication running longer than 10 minutes before the pid is killed and restarted?


#2

@adambethuggin It sounds like you are using the “old” replication worker rather than the new replication configuration.

I’m working on adding some explicit deprecation for the replication worker, but if you’re just looking at replication for the first time, there is a separate UI for configuring “new” replication.

Given the point you are at now, there would be a few steps to get new replication working, you would have to:

  • Disable the role
  • Uninstall replication on the regions (vmdb && bin/rake evm:dbsync:uninstall)
  • Enable new replication on the regions
    • Settings -> Region -> Replication Tab
    • Set type to “Remote”
  • On the master region go to the same tab
    • Set type to “Global”
    • Add dsn information for all regional databases

This route doesn’t use a replication worker and doesn’t require that the role is enabled.

Let me know if this makes sense and if there is anything else I can do.


#3

Hello,

I found this discussion interesting and would have another question.

Will it be possible to setup replication (or synchronization) between several VMDB’s within one region. To set up high available cluster of databases to remove single point of failure?

Thanks and kind regards,
Vaclav


ManageIQ Fine-3 HA in VMware
#4

@vmi

Unfortunately, the current replication feature is not suitable for high-availability use cases.

We are actively working on a user-friendly HA solution and hope to have something like that for the Euwe release.

Nick


#5

Thanks for info Nick!
Good luck in development :slight_smile:


#6

Hi @carbonin,

I’m in the same situation than @vmi , setting up a Master VMDB Cluster (multi-regional architecture) to provide HA to Region 99, but I’m not entire sure if the replication it still will work after a failover. I will setup in both VMDB nodes the ‘Global’ type with the same subscriptors (local cluster vmdb of regions 1, 2, 3 and 4).

Please let me know if you have some info about it.

Thank you. Regards!


#7

@nezf Since this issue was originally opened we added a built-in HA feature which simplifies setting up databases for failover.

The installation process is documented here. That guide also assumes you’ve stopped and disabled the default installation of ManageIQ which starts on first boot of the appliance.

Given all of that, the global region should recover from a failover of the global database without issue. There is an open issue where the region-to-region replication will not recover if a failover occurs in a remote region. You can work around this by removing and re-adding the subscription to the new primary database in the remote region.

I will setup in both VMDB nodes the ‘Global’ type with the same subscriptors

This should not be necessary, all you need is a standby database replicating from the existing global DB, no need to set up the subscriptions again.


#8

If you have more issues regarding this architecture, I would open a new talk topic as this one is quite old at this point.


#9

Hi @carbonin,

Thank you for your time. It will not be necessary because I just had some doubts about the ‘Global’ setup in a master vmdb cluster, but with your explanation was ok. I followed up your guide to setup my environment indeed.

Thanks a lot!


#10

Just the last question… Is there any recommendation about the tables to skip in the Global/Local sync?
Thanks a lot!