Is there a way to limit the number of simultaneous donor/ssts going on in a cluster?
We have the following case that happens to us. Each datacenter has 2 nodes in the following states:
Datacenter A - Synced, Synced
Datacenter B - Synced, Synced
Datacenter C - Synced, Synced
Then we will have a power outage or some other network issue in one of the datacenters resulting in the following:
Datacenter A - Synced, Synced
Datacenter B - Synced, Synced
Datacenter C - Offline, Offline
When datacenter C rejoins the cluster both nodes require a SST resulting in the following state:
Datacenter A - Synced, Synced
Datacenter B - Donor, Donor
Datacenter C - Receiving SST, Receiving SST
Thus, our cluster with 6 nodes drops all the way down to two functional nodes which are not enough to handle the query volume. It also causes a network bottleneck as Datacenter's B & C now are using a lot of bandwidth to make the transfer. Is it possible to limit the cluster to only have 1 SST going on at any particular time? It's better for the nodes to delay coming back up (one at a time) rather than try to SST both at the same time.
We have optimized other settings, such as the size of the galera gcache, etc.
Thank you!