redigo
redigo copied to clipboard
gomodule
add pool FILO mode
There is already a https://github.com/gomodule/redigo/pull/571 for this. The same issue applies to this is that this breaks connection cleanup and is only needed to solve a problem outside of this package.
If I missed how this avoids the issue I’m all ears. However there is also the additional data race if the LIFO variable was ever changed on the fly.
With the current lack of a functional arg constructor, which would require a major version bump, I’m not sure there’s a way to fix 😞
thanks you. We consulted the technical staff of the cloud vendor, and the solution given is this method. I am willing to participate in the follow-up and solve the related problems.
Happy to consider options but from what I’ve read about the situation it would need a more situation aware solution to fix properly.
I wonder if the best fix for this would be to limit the connection lifetime, which would force rebalance over time?
If the lifetime is very short, Pool will be created and closed conn frequently. The use of pool is meaningless and is equivalent to creating a new conn every time. If the time is longer, for example, 20 seconds or 1 minute，Then the rebalan is has no effect，unbalanced state transfers on different connection，single unbalanced connection will still become the overall performance bottleneck。
That would only be the case if the LB isn't doing it's job properly, and if that is the case no fix will work.
From what I understood the balance issue only occurred where there was an issue, hence a 1min max connection lifetime would enable it to only be unbalanced after said issue for 1mins, which would maintain the full benefit of the pool?