percona xtraDB cluster 8.0.19 HA and DR solution across 2x data centers

hi,

we are having a percona xtraDB cluster 5.7x with 2 nodes in primary and another 2 x nodes in DR site, and as mysql can’t failover to a DR site by design we use 2 x ProxySQL on each site to redirect application request to the DR site.

now we will migrate to percona xtraDB cluster 8.0.19 and as 3 x nodes is recommended for a cluster we will use this numbers! can we have 3 x nodes on primary site and 2 x nodes on DR site ? and use the same 2x proxysql in each site to redirect client traffic to any living percona ?

any HA and DR topology for percona XtraDB cluster good reading can teach me this ?

If I understand well for why you want to DR site, maybe try use arbitrator? it will be work like third node.

“maybe try use arbitrator? it will be work like third node.”
yeah, but it don’t store data, right? it just for quorum votes only ?

here every company as DR site, for compliance purpose!  are you saying percona xtraDB cluster is not for cross site ?
we here use percona xtraDB cluster 5.7x across 2 x DC but we use proxySQL to failover. how about  for xtraDB cluster 8.0.19, how we design the 2 x DCs design ? how many nodes in primary and DR site ?  3 x nodes in primary site and 2 x in DR sites and KEEP using proxySQL to failover instead of DB level ?

hi,
any update for me ?

Please explain me, why you want to use DR site? If you have 5 nodes, use all in master PXC. For me it is quite better solution than automatic failover. 

", why you want to use DR site? "
as I said before, here every company as DR site, for compliance purpose! 
“If you have 5 nodes, use all in master PXC”
what is that mean ? all  5 nodes in primary site ? we are now using 2 x nodes in primary site and 2 x nodes in secondary site! in galera replication, if primary site failed, DR site may also stop working as it is even number of nodes !

what I knew is open source DB most likely can’t cross site… 

one thing, as we are using 2 x nodes on primary and another 2 x nodes on DR site, and we use ProxySQL to redirect SQL request to other percona Mysql nodes if one failed. Is it a normal way to architect a Percona XtraDB Cluster 5.7.23 and 8.0.19 DB system ? 

No! if you have at least 2 nodes PXC will be still working. if you lose one of those two nodes then you lose PXC, that why arbiter are cheap solution to secure. 3 nodes in cluster are recommendet because you may lose one of nodes, and your PXC will be works.2 nodes works like 1 node, if you lose 1 you lose all.

explain one thing, your 2xnodes on primary are masters and 2xnodes on DR are slaves to those masters? 

“if you have at least 2 nodes PXC will be still working.”
“2 nodes works like 1 node, if you lose 1 you lose all.”
Yes I know!  then it is not a cluster any more as lost one, the cluster down! not protecting the cluster!
" if you lose one of those two nodes then you lose PXC"

that’s why I prefer 3 nodes. lost 1 nodes PXC (Percona xtraDG cluster, right?) still works , right ?
"that why arbiter are cheap solution to secure. “
yeah, 
”, your 2xnodes on primary are masters and 2xnodes on DR are slaves to those masters? "
1 nodes on primary are master, the rest all slave.

but any configuration  I can use to verify it ? what is the query ?

“1 nodes on primary are master, the rest all slave.”
But this is simple Master-Slave replication, not Percona XtraDB Custer who is Master-Master replication.
https://www.percona.com/doc/percona-xtradb-cluster/LATEST/intro.html