Oh, and one other way to lower your RAM is to scale out….add more machines. Since bloomfilters use up a lot of memory, doubling your cluster and significantly reduce your RAM usage. We have switched to LCS but are being forced to double our cluster as well which reduces RAM quite a bit. Though perhaps like us you are trying to tune to get more per node as well ;). But I thought I would let you know in case it wasn't obvious.
Dean From: Alain RODRIGUEZ <arodr...@gmail.com<mailto:arodr...@gmail.com>> Reply-To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org>> Date: Thursday, March 14, 2013 6:41 AM To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org>> Subject: Re: About the heap "Using half as many m1.xlarge is the way to go." OK, good to know. Are you getting too much GC or running OOM ? GC, it is always gc, I neved had OOM as far as I remember. "Are you using the default GC configuration ?" Yes, as I don't know a lot about it and think default should be fine. Is cassandra logging a lot of GC warnings ? Yes, slowing nodes and even causing a node to be marked down from times to times. I have this kind of message logged in: INFO [ScheduledTasks:1] 2013-03-13 09:10:15,382 GCInspector.java (line 122) GC for ParNew: 212 ms for 1 collections, 4755815744 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-13 09:10:40,000 GCInspector.java (line 122) GC for ParNew: 229 ms for 1 collections, 5432008416 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-13 09:10:41,000 GCInspector.java (line 122) GC for ParNew: 310 ms for 1 collections, 5434752016 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-13 09:10:52,006 GCInspector.java (line 122) GC for ParNew: 215 ms for 1 collections, 5807823960 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-13 09:10:53,007 GCInspector.java (line 122) GC for ParNew: 224 ms for 1 collections, 5765842928 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-13 09:11:18,274 GCInspector.java (line 122) GC for ParNew: 478 ms for 1 collections, 6011120760 used; max is 8547991552 and even this when things goes worst: INFO [ScheduledTasks:1] 2013-03-11 15:02:12,001 GCInspector.java (line 122) GC for ParNew: 626 ms for 1 collections, 7446160296 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-11 15:02:14,002 GCInspector.java (line 122) GC for ParNew: 733 ms for 2 collections, 7777586576 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-11 15:02:15,564 GCInspector.java (line 122) GC for ParNew: 622 ms for 1 collections, 7967657624 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-11 15:02:54,089 GCInspector.java (line 122) GC for ConcurrentMarkSweep: 8460 ms for 2 collections, 7949200768 used; max is 8547991552 WARN [ScheduledTasks:1] 2013-03-11 15:02:54,241 GCInspector.java (line 145) Heap is 0.9299495348869525 full. You may need to reduce memtable and/or cache sizes. Cassandra will now flush up to the two largest memtables to free up memory. Adjust flush_largest_memtables_at threshold in cassandra.yaml if you don't want Cassandra to do this automatically INFO [ScheduledTasks:1] 2013-03-11 15:03:36,487 GCInspector.java (line 122) GC for ConcurrentMarkSweep: 11637 ms for 2 collections, 8367456784 used; max is 8547991552 WARN [ScheduledTasks:1] 2013-03-11 15:03:37,194 GCInspector.java (line 145) Heap is 0.9788798612046171 full. You may need to reduce memtable and/or cache sizes. Cassandra will now flush up to the two largest memtables to free up memory. Adjust flush_largest_memtables_at threshold in cassandra.yaml if you don't want Cassandra to do this automatically INFO [ScheduledTasks:1] 2013-03-11 15:04:19,499 GCInspector.java (line 122) GC for ConcurrentMarkSweep: 11398 ms for 2 collections, 8472967584 used; max is 8547991552 WARN [ScheduledTasks:1] 2013-03-11 15:04:20,096 GCInspector.java (line 145) Heap is 0.9912232051770751 full. You may need to reduce memtable and/or cache sizes. Cassandra will now flush up to the two largest memtables to free up memory. Adjust flush_largest_memtables_at threshold in cassandra.yaml if you don't want Cassandra to do this automatically INFO [ScheduledTasks:1] 2013-03-11 15:05:02,916 GCInspector.java (line 122) GC for ConcurrentMarkSweep: 11877 ms for 2 collections, 8508628816 used; max is 8547991552 WARN [ScheduledTasks:1] 2013-03-11 15:05:02,999 GCInspector.java (line 145) Heap is 0.99539508950605 full. You may need to reduce memtable and/or cache sizes. Cassandra will now flush up to the two largest memtables to free up memory. Adjust flush_largest_memtables_at threshold in cassandra.yaml if you don't want Cassandra to do this automatically INFO [ScheduledTasks:1] 2013-03-11 15:05:42,449 GCInspector.java (line 122) GC for ConcurrentMarkSweep: 11958 ms for 2 collections, 7557641672 used; max is 8547991552 WARN [ScheduledTasks:1] 2013-03-11 15:05:42,813 GCInspector.java (line 145) Heap is 0.8841423890073588 full. You may need to reduce memtable and/or cache sizes. Cassandra will now flush up to the two largest memtables to free up memory. Adjust flush_largest_memtables_at threshold in cassandra.yaml if you don't want Cassandra to do this automatically INFO [ScheduledTasks:1] 2013-03-11 15:05:46,152 GCInspector.java (line 122) GC for ParNew: 665 ms for 1 collections, 8023369408 used; max is 8547991552 INFO [ScheduledTasks:1] 2013-03-11 15:06:18,931 GCInspector.java (line 122) GC for ConcurrentMarkSweep: 9467 ms for 2 collections, 5797092296 used; max is 8547991552 Once again with 1 GB BF, 1GB memtables and 100 MB caches... I am not sure how to avoid this. 2013/3/14 aaron morton <aa...@thelastpickle.com<mailto:aa...@thelastpickle.com>> Because of this I have an unstable cluster and have no other choice than use Amazon EC2 xLarge instances when we would rather use twice more EC2 Large nodes. m1.xlarge is a MUCH better choice than m1.large. You get more ram and better IO and less steal. Using half as many m1.xlarge is the way to go. My heap is actually changing from 3-4 GB to 6 GB and sometimes growing to the max 8 GB (crashing the node). How is it crashing ? Are you getting too much GC or running OOM ? Are you using the default GC configuration ? Is cassandra logging a lot of GC warnings ? If you are running OOM then something has to change. Maybe bloom filters, maybe caches. Enable the GC logging in cassandra-env.sh to check how low a CMS compaction get's the heap, or use some other tool. That will give an idea of how much memory you are using. Here is some background on what is kept on heap in pre 1.2 http://www.mail-archive.com/user@cassandra.apache.org/msg25762.html Cheers ----------------- Aaron Morton Freelance Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 13/03/2013, at 12:19 PM, Wei Zhu <wz1...@yahoo.com<mailto:wz1...@yahoo.com>> wrote: Here is the JIRA I submitted regarding the ancestor. https://issues.apache.org/jira/browse/CASSANDRA-5342 -Wei ----- Original Message ----- From: "Wei Zhu" <wz1...@yahoo.com<mailto:wz1...@yahoo.com>> To: user@cassandra.apache.org<mailto:user@cassandra.apache.org> Sent: Wednesday, March 13, 2013 11:35:29 AM Subject: Re: About the heap Hi Dean, The index_interval is controlling the sampling of the SSTable to speed up the lookup of the keys in the SSTable. Here is the code: https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/db/DataTracker.java#L478 To increase the interval meaning, taking less samples, less memory, slower lookup for read. I did do a heap dump on my production system which caused about 10 seconds pause of the node. I found something interesting, for LCS, it could involve thousands of SSTables for one compaction, the ancestors are recorded in case something goes wrong during the compaction. But those are never removed after the compaction is done. In our case, it takes about 1G of heap memory to store that. I am going to submit a JIRA for that. Here is the culprit: https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/io/sstable/SSTableMetadata.java#L58 Enjoy looking at Cassandra code:) -Wei ----- Original Message ----- From: "Dean Hiller" <dean.hil...@nrel.gov<mailto:dean.hil...@nrel.gov>> To: user@cassandra.apache.org<mailto:user@cassandra.apache.org> Sent: Wednesday, March 13, 2013 11:11:14 AM Subject: Re: About the heap Going to 1.2.2 helped us quite a bit as well as turning on LCS from STCS which gave us smaller bloomfilters. As far as key cache. There is an entry in cassandra.yaml called index_interval set to 128. I am not sure if that is related to key_cache. I think it is. By turning that to 512 or maybe even 1024, you will consume less ram there as well though I ran this test in QA and my key cache size stayed the same so I am really not sure(I am actually checking out cassandra code now to dig a little deeper into this property. Dean From: Alain RODRIGUEZ <arodr...@gmail.com<mailto:arodr...@gmail.com><mailto:arodr...@gmail.com<mailto:arodr...@gmail.com>>> Reply-To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>> Date: Wednesday, March 13, 2013 10:11 AM To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>> Subject: About the heap Hi, I would like to know everything that is in the heap. We are here speaking of C*1.1.6 Theory : - Memtable (1024 MB) - Key Cache (100 MB) - Row Cache (disabled, and serialized with JNA activated anyway, so should be off-heap) - BloomFilters (about 1,03 GB - from cfstats, adding all the "Bloom Filter Space Used" and considering they are showed in Bytes - 1103765112) - Anything else ? So my heap should be fluctuating between 1,15 GB and 2.15 GB and growing slowly (from the new BF of my new data). My heap is actually changing from 3-4 GB to 6 GB and sometimes growing to the max 8 GB (crashing the node). Because of this I have an unstable cluster and have no other choice than use Amazon EC2 xLarge instances when we would rather use twice more EC2 Large nodes. What am I missing ? Practice : Is there a way not inducing any load and easy to do to dump the heap to analyse it with MAT (or anything else that you could advice) ? Alain
