Load balance mode #62
Conversation
| } | ||
| } | ||
|
|
||
| // If we found the last successful client,we'll rotate the list so that the next client is at the beginning |
There was a problem hiding this comment.
| // If we found the last successful client,we'll rotate the list so that the next client is at the beginning | |
| // If we found the last successful client, we'll rotate the list so that the next client is at the beginning |
linglingye001
force-pushed
the
user/linglingye/loadbalance
branch
from
1a38b87 to
3f10c09
Compare
|
|
||
| // If we found the last successful client,we'll rotate the list so that the next client is at the beginning | ||
| if nextClientIndex < len(clients) { | ||
| clients = append(clients[nextClientIndex:], clients[:nextClientIndex]...) |
There was a problem hiding this comment.
I have concern about this, this would be executed in super high frequency, almost every time of executing the ExecuteFailoverPolicy() right? It would create new array every time, have to allocate a new memory space to accommodate the new array.
There was a problem hiding this comment.
We can think of a more effective way to do the array rotation.
There was a problem hiding this comment.
Array rotation is executed every time when ExecuteFailoverPolicy called, I don't think the frequency will be super high. It depends on the refresh interval the customer set and if load balance mode enabled. Besides, the number of replicas usually will not reach hundreds, O(n) time complexity is acceptable.
There was a problem hiding this comment.
But it may have many AzureAppConfigurationProvider yaml resources in one cluster to reconcile, and the improvement would be super easy to implement, so I think it's worth making a change.
No description provided.