You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by Florin Razvan Gheorghe <ra...@gmail.com> on 2015/08/20 17:28:48 UTC

Fwd: JSVC grows until it is killed by the kernel (restarting Cassandra)

Hi all,

I am trying to troubleshoot a case in which the jsvc process (in which
Cassandra is running) uses more and more memory until the OS kills the
process and so Cassandra is restarted.

OS: Ubuntu 12.04
Java: Oracle JDK 8u45
Cassandra: 1.2.19 (datastax)

*Note: We are currently tied to Cassandra 1.2.x due to our implementation
of our services...

Systems:
CPU 4 * 2.3 GHz
Memory 8 GB

Cassandra: out-of-the box parameters, 4 DC with 3 nodes in each DC. 4
keyspaces created by us (one covers all the nodes, 3 of them cover DC1 and
another DC)

Problem: The jsvc memory grows larger and larger over the time and we need
to restart the service every 5-7 days.

Tried the following configuration changes in cassandra:
- memtable_total_space_in_mb and commitlog_total_space_in_mb set to 1024
 - key_cache_size_in_mb: 0

- MAX_HEAP_SIZE="3G" and HEAP_NEWSIZE="200M"
- disk_access_mode: standard

Please let me know your approach in troubleshooting this.

Thank you for your time and patience,
Florin