introduces configurable dispatch hashring replication factor #1227
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
area/CLI
ecordell
reviewed
Mar 28, 2023
| @@ -108,6 +111,10 @@ type Config struct { | |||
| TelemetryInterval time.Duration | |||
| } | |||
|
|
|||
| const defaultBackendsPerKey = 1 | |||
There was a problem hiding this comment.
it would be nice to make this configurable too while we're here
There was a problem hiding this comment.
does that have practical use in SpiceDB? it would cause multiple SpiceDB nodes to have the same key range, which will reduce the likelihood for cache hits, right?
There was a problem hiding this comment.
it's not very useful with smaller clusters, but with larger clusters it lets you spread the load for a hot key between multiple nodes
vroldanbet
force-pushed
the
configurable-replication-factor
branch
from
dd0f1d8
to
47f0722
Compare
so that we get most of the functionality of the server turned on
vroldanbet
force-pushed
the
configurable-replication-factor
branch
from
47f0722
to
5481d3d
Compare
this new CLI allows parameterizing the replication factor used in hashring used by the dispatch client-side gRPC balancer. This was a bit more involved as the balancer registry is a global singleton meant to be registered on an init block. This would have required moving flag handling outside of cobra, which is not ideal. As a solution I changed how the balancer is named, appending the replication factor to the name, and introducing a lazy-load mechanism with a mutex: - if a balancer with a specific replication factor does not exist, lock is acquired and the balancer is registered. This avoids races if multiple spicedb servers are running in the same process - if a balancer exists already, it's not registered - we do not unregister the balancer when the server is wind down, as it could affect another spicedb instance in the process using the same replication factor
vroldanbet
force-pushed
the
configurable-replication-factor
branch
from
5481d3d
to
34a4202
Compare
ecordell
approved these changes
Mar 29, 2023
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
this new CLI allows parameterizing the replication factor
used in hashring used by the dispatch client-side gRPC balancer.
This was a bit more involved as the balancer registry is a global
singleton meant to be registered on an init block. This would have
required moving flag handling outside of cobra, which is not ideal.
As a solution I changed how the balancer is named, appending the replication
factor to the name, and introducing a lazy-load mechanism with a mutex:
acquired and the balancer is registered. This avoids races if multiple
spicedb servers are running in the same process
it could affect another spicedb instance in the process using
the same replication factor