[CASSANDRA-18184] 'Maximum memory usage reached' chunk cache log message doesn't specify which cache is exhausted

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 4.0.8, 4.1.1, 5.0-alpha1, 5.0
• Component/s: Observability/Logging
• Bug Category: Code
• Severity: Normal
• Complexity: Low Hanging Fruit
• Discovered By: User Report
• Platform: All
• Since Version: 4.0-alpha4

Source Control Link

https://github.com/apache/cassandra/commit/0c58fbb8dd25beab4b4a81650be1ed0ec888ff66

Test and Documentation Plan

Submitted CI tasks.

Link to the patch: 18184-trunk.txt

 

• Assignee: Yong Jiang
• Reporter: Brad Schoening

Authors

Yong Jiang

Reviewers

Brandon Williams, Stefan Miklosovic

• Created: 20/Jan/23 19:28
• Updated: 12/Sep/23 13:02
• Resolved: 02/Feb/23 17:49

Description

With Cassandra 4.0.x, we are seeing this cassandra.log message very frequently on the majority of our Cassandra 4.0 clusters:

    [INFO ] [epollEventLoopGroup-5-3] cluster_id=99 ip_address=127.0.0.50  NoSpamLogger.java:92 - Maximum memory usage reached (128.000MiB), cannot allocate chunk of 8.000MiB

It took me several weeks to track down what it means, until I saw this start-up message

    BufferPools.java:49 - Global buffer pool limit is 2.000GiB for chunk-cache and 128.000MiB for networking

This maximum memory usage warning would benefit from clarifying that its the network cache, not the off-heap chunk cache which is exhausted.  With 'chunk cache' in both warning messages, they're too easily confused.