Made akka.cluster.sharding.verbose-debug-logging accept a LogLevel
#7118
Conversation
| @@ -28,7 +28,7 @@ public class ClusterShardingLeavingSpecConfig : MultiNodeClusterShardingConfig | |||
|
|
|||
| public ClusterShardingLeavingSpecConfig(StateStoreMode mode) | |||
| : base(mode: mode, loglevel: "DEBUG", additionalConfig: @" | |||
| akka.cluster.sharding.verbose-debug-logging = on | |||
| akka.cluster.sharding.verbose-debug-logging = INFO | |||
There was a problem hiding this comment.
Added a test case to ensure that this actually works with INFO logging here. Going to inspect the output once this runs.
| @@ -29,7 +30,7 @@ internal sealed class DDataShardCoordinator : ActorBase, IWithTimers, IWithUnbou | |||
|
|
|||
| private sealed class RememberEntitiesStoreStopped | |||
| { | |||
| public static RememberEntitiesStoreStopped Instance = new(); | |||
| public static readonly RememberEntitiesStoreStopped Instance = new(); | |||
There was a problem hiding this comment.
Minor fix unrelated to this PR, but noticed it while I was working.
| @@ -77,7 +78,7 @@ private RememberEntitiesLoadTimeout() | |||
|
|
|||
| private readonly IActorRef _replicator; | |||
| private readonly ShardCoordinator _baseImpl; | |||
| private bool VerboseDebug => _baseImpl.VerboseDebug; | |||
| private LogLevel? VerboseDebug => _baseImpl.VerboseDebug; | |||
There was a problem hiding this comment.
Change from a bool to a LogLevel? - these are all private APIs so there's no user-facing impact here other than the logging output.
| @@ -108,7 +109,7 @@ private RememberEntitiesLoadTimeout() | |||
| { | |||
| _replicator = replicator; | |||
| var log = Context.GetLogger(); | |||
| var verboseDebug = Context.System.Settings.Config.GetBoolean("akka.cluster.sharding.verbose-debug-logging"); | |||
| var verboseDebug = ParseVerboseLogSettings(Context.System.Settings.Config); | |||
There was a problem hiding this comment.
Added a utility method to parse these configuration settings into LogLevel? since we have to make this call from both the DDataShardCoordinator and the PersistentShardCoordinator.
| if (VerboseDebug) | ||
| Log.Debug("{0}: Received initial coordinator state [{1}]", TypeName, existingState); | ||
| else | ||
| if (VerboseDebug.HasValue) |
There was a problem hiding this comment.
The way I handled this type of case, which came up in a few areas:
- In the past we use to have two different
Debugcalls - one with more data whenVerboseDebuggingwas enabled and then a regular one that was used withLog.IsDebugEnabledwas true. - All of these calls have been converted to include the first call to use whatever
LogLevelis specified byVerboseDebuggingin the event thatVerboseDebuggingis non-null. - If
VerboseDebuggingisnullandLog.IsDebugEnabled, we record the "lighter" information.
There was a problem hiding this comment.
Internal utility class for parsing the LogLevel?
| # in production systems. | ||
| # in production systems. | ||
| # Values can be `off`, `on`, or an `akka.loglevel` value such as `INFO`. | ||
| # When set to `on` these logs will be logged using `LogLevel.Debug`. |
There was a problem hiding this comment.
New comments describing the underlying HOCON changes. The default value of the verbose-debug-logging setting has been left intact.
| "off" => null, | ||
| "on" => LogLevel.DebugLevel, |
There was a problem hiding this comment.
To conform to HOCON standard, this will also need to check for "true" and "false" value.
There was a problem hiding this comment.
Good point - will do.
I think this comes down to that our log sources aren't easily filterable on the back-end, so it has to happen on the front-end. This change is also pretty easy to reverse if there were an issue with it. |
I actually have some ideas on how that can be addressed too, in the grand scheme of things, but that'd be a 1.6 or 2.0 feature. |
Changes
close #7097
For users with a large number of sharded entities, the
akka.cluster.sharding.verbose-debug-loggingHOCON setting now accepts aLogLevelargument. It also still supports the same boolean values as before too.However, if you were to pass in the following HOCON:
All of the internal
ShardCoordinatorlogs could now be captured at aLogLevel.Infolevel, so there wouldn't be any need to turnLogLevel.Debugon globally to capture these logs.Checklist
For significant changes, please ensure that the following have been completed (delete if not relevant):