Even with the same data, bloom filter is based on sstables. If your compaction behaves differently on 2 nodes than the third, your bloom filter RAM usage may be different.
From: Kai Wang Reply-To: "user@cassandra.apache.org" Date: Tuesday, May 17, 2016 at 8:02 PM To: "user@cassandra.apache.org" Subject: Re: Bloom filter memory usage disparity Alain, Thanks for replying. I am using C* 2.2.4. Yes the table is RF=3. I changed bloom_filter_fp_chance from 0.01 to 0.1 a couple of months ago. On Tue, May 17, 2016 at 11:05 AM, Alain RODRIGUEZ <arodr...@gmail.com> wrote: Hi, we would need more information here (if you did not solve it yet). What is your Cassandra version? Does this 3 node cluster use a Replication Factor of 3? Did you change the bloom_filter_fp_chance recently? That table has about 16M keys and 140GB of data. Is that the total value or per node? In any case, we need the data size for the 3 nodes to understand. It might have been a temporary situation, but in this case you would know by now. C*heers, 2016-05-03 18:47 GMT+02:00 Kai Wang <dep...@gmail.com>: Hi, I have a table on 3-node cluster. I notice bloom filter memory usage are very different on one of the node. For a given table, I checked CassandraMetricsRegistry$JmxGauge.[table]_BloomFilterOffHeapMemoryUsed.Value. 2 of 3 nodes show 1.5GB while the other shows 2.5 GB. What could be the reason? That table is using LCS. bloom_filter_fp_chance=0.1 That table has about 16M keys and 140GB of data. Thanks.
smime.p7s
Description: S/MIME cryptographic signature