Skip to content

[ASSIGNOR] 調查端對端延遲過高原因 #1774

@harryteng9527

Description

@harryteng9527

related #1524 (comment)

此 issue 拿來討論實驗中使用 Cost-Aware assignor 的 consumers 端對端延遲有時候會到 20 多秒

Partition Lag

下圖 partition lag 的圖是抓 Kafka consumer 紀錄的 Lag 所繪製的圖
partitionLAg

上面 partition lag 的圖中,使用 Cost-Aware assignor 的 consumers 會追不上兩個 partitions 的 Lag,而使用 Range assignor 的 consumers 追不上的 partitions 數量更多

我去看了一下 Cost-Aware 中 consumers 的端對端延遲,看到被分配到這兩個 partitions 的 consumer 的端對端延遲有時會特別高(下圖的 Consumer[1] 70)
Consumers 的端對端延遲

看它(Consumer[1] 70)的分配沒有被流入速率低的 partition 影響,但被分配的 cost 是最重的,比平均高 50 MiB,所以可能是來不及處理導致資料一直堆積,產生過高的 e2e latency

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions