wsrep_streaming_log size not changing

[Can Tokay]

Hello,

I have some questions about galera-4 with mariadb 10.4

I am using streaming replication in order to import a huge database. Even if i completed my sqldump and drop my database, wsrep_streaming_log.ibd still seems huge. How can i flush this table? Also, wsrep_flow_control_paused value is decreasing but too slow. What is the meaning of this value?

Could you please help me?

Regards.

[rey wang]

Is this setup with AWS multi-zones , or on-prem and how big of the database (in Gig Bytes) ?

What is your daily new data changes (KBytes) ?

[Can Tokay]

Hello

This setup is on my vmware cluster. Database is nearly 100GB. I imported my database using SQLYog then observe what is happenning. 

Because of the commit error on the previous version of galera, I am using streaming replication with galera-4.  Everything is fine now, however wsrep_streaming_log size table is nearly 70GB and when i drop my database, this table size stucking at the same size. 

One more thing; wsrep_flow_control_paused value is decreasing continuously but what is that meaning.? 

Thanks for your reply.

Regards.

[rey wang]

SQLyog is windows app ; is your mariaDB cluster Linux / Ubuntu OS ?

By any chance, your step also need to issue GALERA_NEW_CLUSTER(1)? 

[Can Tokay]

Thanks for your reply.

I have 3 Centos 7 VM. I am using galera-4 and mariadb 10.4 on them.

I am using SQLyog only for copying database. Because i don't want to dump an sql file and then import them.I am just want use it for copy operation.

By the way i have experience about galera cluster. This 3 vm is already working as cluster. no need to issue galera_new_cluster command. My question is not about creating new cluster. I want to understand how can i reduce wsrep_streaming_log size  or when it changes? i mean it is very big like 60GB.?  Also i want to understand wsrep_flow_control_paused. It is decreasing but too slow. And if i restart one node of my cluster while this value decreasing, restarted node will not come back again.

Regards.