We need to investigate, why sometimes router fails to send requests to instance during rebalancing process. We should add tests in vshard, that under load all requests successfully pass. Or at least only several requests fail. The number of failed requests must be minimized.
On the first glance, I cannot see reason for that, as router always try to update its routing table and retry request, when bucket is moved.
Currently vshard requests rarely fail during rebalancing, which affects users in mission critical projects.
We need to investigate, why sometimes router fails to send requests to instance during rebalancing process. We should add tests in vshard, that under load all requests successfully pass. Or at least only several requests fail. The number of failed requests must be minimized.
On the first glance, I cannot see reason for that, as router always try to update its routing table and retry request, when bucket is moved.
Currently vshard requests rarely fail during rebalancing, which affects users in mission critical projects.