You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
10ed356 add e2e feature documentation, updated readme with link
ec428c0 fix issues when with logging string pointers
390cf6f multiple changes: - remove rebalancingProtocol config, since there will never be any rebalancing in kminions consumer groups (since every kminion instance has exactly one, unique, consumer group that it uses) - change type of 'requiredAcks' from int to string (now takes 'all' or 'leader') - lots of functions have been refactored to use "early return" and/or split into logical steps so there aren't any huge functions - 'createEndToEndRecord' doesn't return an error anymore and instead panics. if serialization fails, there must be an issue that prevents kminion from running anyway - actually set 'ProduceRequestTimeout' to 'ackSla' config property - topic.go: complete rewrite: properly validate partition count, partition assignments, replication factor; considering potentially changing broker/partition count and replicationFactor; also now error checking for all potential errors in kafka responses (including nested errors, that are for example local to only one partition)
e262749 Merge pull request #90 from JoschaLaubach/fix-link-in-charts-docs
c2df9de rename "image-on-push" to "image-on-push-to-master", and create a new workflow that only builds images for pushes to any other (non-master) branch and tagging the image as "branchName-gitSha"
e1a8788 put deleting old consumer groups into its own file (group_tracker.go)
e36eb71 each kminion instance uses its own consumer group for end-to-end now so the instances don't "steal" messages from each other
d5a116c remove bytesSent/Received metrics from end to end
b4a08c5 more refactoring, seperate client hooks for each package
1349666 move end-to-end into its own package called e2e
cc9a4e9 - remove SLA gauge metrics, they can be inferred from the 'inifinite' bucket in the latency histograms - methods to handle metric events (onAck, onRoundtrip, onOffsetCommit); will also soon be used to extract e2e into its own package - add _total and _seconds suffixes to metrics for best practices
9813ad2 remove old methods; extract 'processMessage' for a small cleanup
c8e360a many changes: - new metrics (in 3 groups): low level, SLA indicators, latency histograms - don't put metrics results in a cache, instead directly increment/observe/set on promauto objects - only generate minionID once, don't create a new one for every message - actually check minionID to filter messages from other kminion instances - check timestamp to discard old messages and establish roundtrip latency - keep track of last received message timestamp, so we can detect/skip cases where we receive older messages for whatever reason - don't call storage.markRecordConsumed since that's only for offset_consumer