[jira] [Commented] (CASSANDRA-8066) High Heap Consumption due to high number of SSTableReader
[ https://issues.apache.org/jira/browse/CASSANDRA-8066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14169722#comment-14169722 ] Benoit Lacelle commented on CASSANDRA-8066: --- I have a single node. The schema holds around 8 tables. A few of them are counter tables. I regularly run nearly-full scan on on a not-counter table. Else, it is mainly an equivalent of reads and writes on 2 given tables. These tables have a few million of rows. Do you need more details? High Heap Consumption due to high number of SSTableReader - Key: CASSANDRA-8066 URL: https://issues.apache.org/jira/browse/CASSANDRA-8066 Project: Cassandra Issue Type: Bug Components: Core Environment: Cassandra 2.1.0 Reporter: Benoit Lacelle Assignee: T Jake Luciani Fix For: 2.1.1 Given a workload with quite a lot of reads, I recently encountered high heap memory consumption. Given 2GB of Heap, it appears I have 750.000+ tasks in SSTableReader.syncExecutor, consuming more than 1.2GB. These tasks have type SSTableReader$5, which I guess corresponds to : {code} readMeterSyncFuture = syncExecutor.scheduleAtFixedRate(new Runnable() { public void run() { if (!isCompacted.get()) { meterSyncThrottle.acquire(); SystemKeyspace.persistSSTableReadMeter(desc.ksname, desc.cfname, desc.generation, readMeter); } } }, 1, 5, TimeUnit.MINUTES); {code} I do not have have to the environment right now, but I could provide a threaddump later if necessary. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8066) High Heap Consumption due to high number of SSTableReader
[ https://issues.apache.org/jira/browse/CASSANDRA-8066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14169740#comment-14169740 ] T Jake Luciani commented on CASSANDRA-8066: --- bq. Do you need more details? Yes actual data size and number of sstables in the largest keyspace High Heap Consumption due to high number of SSTableReader - Key: CASSANDRA-8066 URL: https://issues.apache.org/jira/browse/CASSANDRA-8066 Project: Cassandra Issue Type: Bug Components: Core Environment: Cassandra 2.1.0 Reporter: Benoit Lacelle Assignee: T Jake Luciani Fix For: 2.1.1 Given a workload with quite a lot of reads, I recently encountered high heap memory consumption. Given 2GB of Heap, it appears I have 750.000+ tasks in SSTableReader.syncExecutor, consuming more than 1.2GB. These tasks have type SSTableReader$5, which I guess corresponds to : {code} readMeterSyncFuture = syncExecutor.scheduleAtFixedRate(new Runnable() { public void run() { if (!isCompacted.get()) { meterSyncThrottle.acquire(); SystemKeyspace.persistSSTableReadMeter(desc.ksname, desc.cfname, desc.generation, readMeter); } } }, 1, 5, TimeUnit.MINUTES); {code} I do not have have to the environment right now, but I could provide a threaddump later if necessary. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8066) High Heap Consumption due to high number of SSTableReader
[ https://issues.apache.org/jira/browse/CASSANDRA-8066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14169775#comment-14169775 ] Benoit Lacelle commented on CASSANDRA-8066: --- Here is the output for nodetool cfstats {code} Keyspace: prod_7 Read Count: 1424130 Read Latency: 3.122707294980093 ms. Write Count: 8808265 Write Latency: 0.03234213866181365 ms. Pending Flushes: 0 Table: alerts SSTable count: 2 Space used (live), bytes: 10237 Space used (total), bytes: 10237 Space used by snapshots (total), bytes: 0 SSTable Compression Ratio: 0.6354418105671432 Memtable cell count: 0 Memtable data size, bytes: 0 Memtable switch count: 0 Local read count: 0 Local read latency: NaN ms Local write count: 0 Local write latency: NaN ms Pending flushes: 0 Bloom filter false positives: 0 Bloom filter false ratio: 0.0 Bloom filter space used, bytes: 32 Compacted partition minimum bytes: 259 Compacted partition maximum bytes: 372 Compacted partition mean bytes: 341 Average live cells per slice (last five minutes): 0.0 Average tombstones per slice (last five minutes): 0.0 Table: details SSTable count: 103 Space used (live), bytes: 578266489 Space used (total), bytes: 578266489 Space used by snapshots (total), bytes: 0 SSTable Compression Ratio: 0.724988344517149 Memtable cell count: 67212 Memtable data size, bytes: 18468770 Memtable switch count: 23 Local read count: 6 Local read latency: 10.742 ms Local write count: 2036971 Local write latency: 0.017 ms Pending flushes: 0 Bloom filter false positives: 0 Bloom filter false ratio: 0.0 Bloom filter space used, bytes: 5136 Compacted partition minimum bytes: 87 Compacted partition maximum bytes: 129557750 Compacted partition mean bytes: 2595076 Average live cells per slice (last five minutes): 0. Average tombstones per slice (last five minutes): 0.0 Table: domains SSTable count: 21 Space used (live), bytes: 122407 Space used (total), bytes: 122407 Space used by snapshots (total), bytes: 0 SSTable Compression Ratio: 0.5848437821906775 Memtable cell count: 238238 Memtable data size, bytes: 2793 Memtable switch count: 1 Local read count: 60281 Local read latency: 0.162 ms Local write count: 402903 Local write latency: 0.012 ms Pending flushes: 0 Bloom filter false positives: 25 Bloom filter false ratio: 0.08929 Bloom filter space used, bytes: 664 Compacted partition minimum bytes: 87 Compacted partition maximum bytes: 372 Compacted partition mean bytes: 171 Average live cells per slice (last five minutes): 0.9985401459854014 Average tombstones per slice (last five minutes): 0.0 Table: domains_statistics SSTable count: 7 Space used (live), bytes: 302413 Space used (total), bytes: 302413 Space used by snapshots (total), bytes: 0 SSTable Compression Ratio: 0.6144160676068052 Memtable cell count: 849893 Memtable data size, bytes: 42569 Memtable switch count: 1 Local read count: 60511 Local read latency: 0.141 ms Local write count: 1055892 Local write latency: 0.013 ms Pending flushes: 0 Bloom filter false positives: 0 Bloom filter false ratio: 0.0 Bloom filter space used, bytes: 416 Compacted partition minimum bytes: 87 Compacted partition maximum bytes: 24601 Compacted partition mean bytes: 3795 Average live cells per slice (last five minutes): 0.9894160583941606 Average tombstones per slice (last five minutes): 0.0 Table: latest_urls SSTable count: 3 Space used (live), bytes: 10239518
[jira] [Commented] (CASSANDRA-8066) High Heap Consumption due to high number of SSTableReader
[ https://issues.apache.org/jira/browse/CASSANDRA-8066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14169938#comment-14169938 ] Jason Brown commented on CASSANDRA-8066: +1 High Heap Consumption due to high number of SSTableReader - Key: CASSANDRA-8066 URL: https://issues.apache.org/jira/browse/CASSANDRA-8066 Project: Cassandra Issue Type: Bug Components: Core Environment: Cassandra 2.1.0 Reporter: Benoit Lacelle Assignee: T Jake Luciani Fix For: 2.1.1 Attachments: 8066.txt Given a workload with quite a lot of reads, I recently encountered high heap memory consumption. Given 2GB of Heap, it appears I have 750.000+ tasks in SSTableReader.syncExecutor, consuming more than 1.2GB. These tasks have type SSTableReader$5, which I guess corresponds to : {code} readMeterSyncFuture = syncExecutor.scheduleAtFixedRate(new Runnable() { public void run() { if (!isCompacted.get()) { meterSyncThrottle.acquire(); SystemKeyspace.persistSSTableReadMeter(desc.ksname, desc.cfname, desc.generation, readMeter); } } }, 1, 5, TimeUnit.MINUTES); {code} I do not have have to the environment right now, but I could provide a threaddump later if necessary. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8066) High Heap Consumption due to high number of SSTableReader
[ https://issues.apache.org/jira/browse/CASSANDRA-8066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14165441#comment-14165441 ] T Jake Luciani commented on CASSANDRA-8066: --- I suspect the early opening and re-opening of sstables during compaction is creating many entries in the queue. We should perhaps postpone this read meter tracking until the sstable has been fully written. High Heap Consumption due to high number of SSTableReader - Key: CASSANDRA-8066 URL: https://issues.apache.org/jira/browse/CASSANDRA-8066 Project: Cassandra Issue Type: Bug Components: Core Environment: Cassandra 2.1.0 Reporter: Benoit Lacelle Assignee: T Jake Luciani Fix For: 2.1.1 Given a workload with quite a lot of reads, I recently encountered high heap memory consumption. Given 2GB of Heap, it appears I have 750.000+ tasks in SSTableReader.syncExecutor, consuming more than 1.2GB. These tasks have type SSTableReader$5, which I guess corresponds to : {code} readMeterSyncFuture = syncExecutor.scheduleAtFixedRate(new Runnable() { public void run() { if (!isCompacted.get()) { meterSyncThrottle.acquire(); SystemKeyspace.persistSSTableReadMeter(desc.ksname, desc.cfname, desc.generation, readMeter); } } }, 1, 5, TimeUnit.MINUTES); {code} I do not have have to the environment right now, but I could provide a threaddump later if necessary. -- This message was sent by Atlassian JIRA (v6.3.4#6332)