The subscriber count is off here as well. I’ve had to subscribe twice and on the second time I notice the subscriber count did not change. So I unsubscribed to watch the number and resubscribed and still no increase or decrease in count. No big deal but thought it was important to state.
you know, if this site is being ran on clustered servers and is using things like Redis cache for tracking stateful numbers like subscriber counts, you could end up seeing different results from one refresh to another based on a load balancer hitting a different cluster node.
in a distributed system, it's typical that a lot of state is tracked in a way that does not comply with the semantics of ACID OLTP transaction semantics - it allows such architectures to scale and typically its not vital to track a lot of the state with ACID OLTP transaction semantics compliance
So there are multiple servers that are kept in sync, but it's expensive to ensure that the synchronization is always completed before data is displayed. Things like subscriber counts may be tracked correctly, but it's not all that important to ensure that you always see a correct current value?
Yes, that's it - in Wikipedia can read about something called CAPS theorem that revolves around this subject matter pertaining to distributed systems architecture
view the rest of the comments →
0x75 ago
The subscriber count is off here as well. I’ve had to subscribe twice and on the second time I notice the subscriber count did not change. So I unsubscribed to watch the number and resubscribed and still no increase or decrease in count. No big deal but thought it was important to state.
Christosgnosis ago
you know, if this site is being ran on clustered servers and is using things like Redis cache for tracking stateful numbers like subscriber counts, you could end up seeing different results from one refresh to another based on a load balancer hitting a different cluster node.
in a distributed system, it's typical that a lot of state is tracked in a way that does not comply with the semantics of ACID OLTP transaction semantics - it allows such architectures to scale and typically its not vital to track a lot of the state with ACID OLTP transaction semantics compliance
hope that's helpful
NickDixon ago
So there are multiple servers that are kept in sync, but it's expensive to ensure that the synchronization is always completed before data is displayed. Things like subscriber counts may be tracked correctly, but it's not all that important to ensure that you always see a correct current value?
Christosgnosis ago
Yes, that's it - in Wikipedia can read about something called CAPS theorem that revolves around this subject matter pertaining to distributed systems architecture