Group replication slow down existing asynchronous replication

MySQL & MariaDB
1

I am running group replication using 8.0.35-27 and this cluster receive transactions from outside using regular asynchronous replication.
The setup described there: https://bugs.mysql.com/114072
I’ve noticed that once latency of network increase between members the asynchronous replication accumulating lag till I am stopping all secondary members.
I am curious if any one else had similar issue.

Thanks
Evgeni

2

Thank you for contacting us,

I asked my colleagues and they don’t seem to have encountered this issue.

As I see you already have a discussion going on in the MySQL bug tracker, hopefully the problem will be resolved.

3

Thank you very much for your time.

No, the issue is not resolved.
We, most probably, would avoid GR over WAN as of now.

Best Regards

4

Hi @egel thank you for sharing your use case with us here on the Forum
Have you deployed any monitoring software against this environment? If not may I suggest PMM as it ships with GR support whereby we can review the throughput and latency between nodes to better understand your issue

Yes absolutely, you should NOT be running your GR members across a WAN. It will “work”, however since GR is virtually synchronous, each write requires cluster-wide communications which can get expensive over slow links.
Best Practices are to run your GR members in the same Region if you can reliably count on fast inter-availability zone communications.

1 Like
5

Yes, but galera based PXC working just fine in the same WAN. Therefore I was surprised to see such a issue with GR.

Thanks

6

Thank you for the feedback @egel ! Marking this topic as solved, we can re-visit this topic if you can share some graphs from PMM

1 Like