cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "T Jake Luciani (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-12585) Fixup Cassandra Stress reporting thread model and precision
Date Tue, 06 Sep 2016 14:18:20 GMT


T Jake Luciani commented on CASSANDRA-12585:


> Fixup Cassandra Stress reporting thread model and precision
> -----------------------------------------------------------
>                 Key: CASSANDRA-12585
>                 URL:
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Tools
>            Reporter: Nitsan Wakart
>            Assignee: Nitsan Wakart
>            Priority: Minor
>              Labels: stress
> Stress current reporting thread currently has a complex, slow and imprecise method of
collecting measurements from the Consumer threads.
> StressMetrics thread iterates through the Timers (one Timer per op per Consumer thread),
blocks until the consumer thread collects a measurement on that operation, which forces the
consumer thread to collect a report and unblock the StressMetrics thread. This means:
> - Reporting intervals are effected by sleep interval accuracy on the StressMetrics thread
> - Interval report timing is staggered as the operations get collected.
> - The intended logging interval often does not resemble the effective interval.
> This patch simplifies collection method and resolves the above issues:
> - Each Consumer thread submits timing data to a queue
> - StressMetrics thread reads from the queues and merges the data into a report
> - StressMetrics reads and reports for the intended interval, so not effected by timing
> - Measurement intervals are aligned to the second, improving the ease of measurements
correlation from several nodes.
> See branch here:

This message was sent by Atlassian JIRA

View raw message