[jira] [Commented] (CASSANDRA-10680) Deal with small compression chunk size better during streaming plan setup
[ https://issues.apache.org/jira/browse/CASSANDRA-10680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15009233#comment-15009233 ] Paulo Motta commented on CASSANDRA-10680: - Patch and tests look good. Thanks! > Deal with small compression chunk size better during streaming plan setup > - > > Key: CASSANDRA-10680 > URL: https://issues.apache.org/jira/browse/CASSANDRA-10680 > Project: Cassandra > Issue Type: Bug >Reporter: Jeff Jirsa >Assignee: Yuki Morishita > Fix For: 3.0.1, 3.1, 2.1.x, 2.2.x > > > For clusters using small compression chunk size and terabytes of data, the > streaming plan calculations will instantiate hundreds of millions of > compressionmetadata$chunk objects, which will create unreasonable amounts of > heap pressure. Rather than instantiating all of those at once, streaming > should instantiate only as many as needed for a single file per table at a > time. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (CASSANDRA-10680) Deal with small compression chunk size better during streaming plan setup
[ https://issues.apache.org/jira/browse/CASSANDRA-10680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15001251#comment-15001251 ] Yuki Morishita commented on CASSANDRA-10680: Patch is ready for review. | 2.1 | 2.2 | 3.0.x (3.1) | | [branch|https://github.com/yukim/cassandra/tree/10680] | [branch|https://github.com/yukim/cassandra/tree/10680-2.2] | [branch|https://github.com/yukim/cassandra/tree/10680-3.0] | | [testall|http://cassci.datastax.com/view/Dev/view/yukim/job/yukim-10680-testall/lastCompletedBuild/testReport/] | [testall|http://cassci.datastax.com/view/Dev/view/yukim/job/yukim-10680-2.2-testall/lastCompletedBuild/testReport/] | [testall|http://cassci.datastax.com/view/Dev/view/yukim/job/yukim-10680-3.0-testall/lastCompletedBuild/testReport/] | | [dtest|http://cassci.datastax.com/view/Dev/view/yukim/job/yukim-10680-dtest/lastCompletedBuild/testReport/] | [dtest|http://cassci.datastax.com/view/Dev/view/yukim/job/yukim-10680-2.2-dtest/lastCompletedBuild/testReport/] | [dtest|http://cassci.datastax.com/view/Dev/view/yukim/job/yukim-10680-2.2-dtest/lastCompletedBuild/testReport/] | > Deal with small compression chunk size better during streaming plan setup > - > > Key: CASSANDRA-10680 > URL: https://issues.apache.org/jira/browse/CASSANDRA-10680 > Project: Cassandra > Issue Type: Bug >Reporter: Jeff Jirsa >Assignee: Yuki Morishita > Fix For: 2.1.x > > > For clusters using small compression chunk size and terabytes of data, the > streaming plan calculations will instantiate hundreds of millions of > compressionmetadata$chunk objects, which will create unreasonable amounts of > heap pressure. Rather than instantiating all of those at once, streaming > should instantiate only as many as needed for a single file per table at a > time. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (CASSANDRA-10680) Deal with small compression chunk size better during streaming plan setup
[ https://issues.apache.org/jira/browse/CASSANDRA-10680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14999459#comment-14999459 ] Jeff Jirsa commented on CASSANDRA-10680: Initial tests suggest this (commit 61d2630e9950e9abc0d8da3939b280ff44b5ddc0) does indeed solve this issue. > Deal with small compression chunk size better during streaming plan setup > - > > Key: CASSANDRA-10680 > URL: https://issues.apache.org/jira/browse/CASSANDRA-10680 > Project: Cassandra > Issue Type: Bug >Reporter: Jeff Jirsa >Assignee: Yuki Morishita > Fix For: 2.1.x > > > For clusters using small compression chunk size and terabytes of data, the > streaming plan calculations will instantiate hundreds of millions of > compressionmetadata$chunk objects, which will create unreasonable amounts of > heap pressure. Rather than instantiating all of those at once, streaming > should instantiate only as many as needed for a single file per table at a > time. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (CASSANDRA-10680) Deal with small compression chunk size better during streaming plan setup
[ https://issues.apache.org/jira/browse/CASSANDRA-10680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14998038#comment-14998038 ] Yuki Morishita commented on CASSANDRA-10680: Patch here: https://github.com/yukim/cassandra/tree/10680 (tests are running.) The patch creates chunks right before sending file instead of the time when adding SSTables to stream. > Deal with small compression chunk size better during streaming plan setup > - > > Key: CASSANDRA-10680 > URL: https://issues.apache.org/jira/browse/CASSANDRA-10680 > Project: Cassandra > Issue Type: Bug >Reporter: Jeff Jirsa >Assignee: Yuki Morishita > Fix For: 2.1.x > > > For clusters using small compression chunk size and terabytes of data, the > streaming plan calculations will instantiate hundreds of millions of > compressionmetadata$chunk objects, which will create unreasonable amounts of > heap pressure. Rather than instantiating all of those at once, streaming > should instantiate only as many as needed for a single file per table at a > time. -- This message was sent by Atlassian JIRA (v6.3.4#6332)