-
Notifications
You must be signed in to change notification settings - Fork 3.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Cluster sharding state warmup with health check #27168
Comments
I think this would be really helpful in elastic environments, thanks for raising! |
Is there any progress on this? Since we deploy on ECS (which shuts down the oldest containers first), we have a lot of movement of our ShardCoordinator during a redeploy. |
One way would be to reply with all shard locations when the region register to the coordinator. I think it's better to add a separate message for this. Then it can maybe be used in other situations like #29589 It could be many shards, and might not fit in one single message, but we can split it several smaller. |
The health check was added in #29638 |
…#27168 This should shorten latencies per new shard that is addressed from a newly joined region/proxy
This should shorten latencies per new shard that is addressed from a newly joined region/proxy
When a new node is added to a cluster using cluster sharding, if user requests are initially routed to it then each of them will incur a latency penalty as the
ShardRegion
gets theShard
location from theShardCoordinator
. This is for allShard
s, not just the ones that will be moved.The
ShardCoordinator
could instead send the current regions per shardId to theShardRegion
when it registers and a health check could be exposed (https://doc.akka.io/docs/akka-management/current/healthchecks.html) to prevent user traffic being routed to the node in environments like Kubernetes until this has happened.This might help #30315.
/cc @jroper
The text was updated successfully, but these errors were encountered: