[jira] [Commented] (NUTCH-2179) Cleanup job for SOLR Performance Boost
[ https://issues.apache.org/jira/browse/NUTCH-2179?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16362912#comment-16362912 ] David Johnson commented on NUTCH-2179: -- Appears to have been rsolved by NUTCH-2197, which introduced a similar delete queue > Cleanup job for SOLR Performance Boost > -- > > Key: NUTCH-2179 > URL: https://issues.apache.org/jira/browse/NUTCH-2179 > Project: Nutch > Issue Type: Improvement > Components: indexer >Affects Versions: 1.9, 1.10, 1.11 >Reporter: David Johnson >Priority: Minor > Labels: patch > Attachments: 0001-Create-delete-queue.patch > > Original Estimate: 2h > Remaining Estimate: 2h > > During a cleanup job, index deletes are scheduled one by one, which can make > a large job take days -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (NUTCH-2179) Cleanup job for SOLR Performance Boost
[ https://issues.apache.org/jira/browse/NUTCH-2179?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15034541#comment-15034541 ] David Johnson commented on NUTCH-2179: -- Correct, it should be, but is currently writing them singleton. I'm proposing we swap the current delete function public void delete(String key) throws IOException { if (delete) { try { solr.deleteById(key); //singleton delete numDeletes++; } catch (final SolrServerException e) { throw makeIOException(e); } } } with public void delete(String key) throws IOException { if (delete) { deleteURLs.add(key); if (inputDocs.size() + deleteURLs.size() >= batchSize) { sendRequest(); } } } and abstract the actual send request with a function that can be called from write, delete and close. I have a patch prepared, but I cannot get the branch to commit - I'll attempt again later. > Cleanup job for SOLR Performance Boost > -- > > Key: NUTCH-2179 > URL: https://issues.apache.org/jira/browse/NUTCH-2179 > Project: Nutch > Issue Type: Improvement > Components: indexer >Affects Versions: 1.9, 1.10, 1.11 >Reporter: David Johnson >Priority: Minor > Labels: patch > Original Estimate: 2h > Remaining Estimate: 2h > > During a cleanup job, index deletes are scheduled one by one, which can make > a large job take days -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (NUTCH-2179) Cleanup job for SOLR Performance Boost
[ https://issues.apache.org/jira/browse/NUTCH-2179?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15034542#comment-15034542 ] David Johnson commented on NUTCH-2179: -- Correct, it should be, but is currently writing them singleton. I'm proposing we swap the current delete function public void delete(String key) throws IOException { if (delete) { try { solr.deleteById(key); //singleton delete numDeletes++; } catch (final SolrServerException e) { throw makeIOException(e); } } } with public void delete(String key) throws IOException { if (delete) { deleteURLs.add(key); if (inputDocs.size() + deleteURLs.size() >= batchSize) { sendRequest(); } } } and abstract the actual send request with a function that can be called from write, delete and close. I have a patch prepared, but I cannot get the branch to commit - I'll attempt again later. > Cleanup job for SOLR Performance Boost > -- > > Key: NUTCH-2179 > URL: https://issues.apache.org/jira/browse/NUTCH-2179 > Project: Nutch > Issue Type: Improvement > Components: indexer >Affects Versions: 1.9, 1.10, 1.11 >Reporter: David Johnson >Priority: Minor > Labels: patch > Original Estimate: 2h > Remaining Estimate: 2h > > During a cleanup job, index deletes are scheduled one by one, which can make > a large job take days -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (NUTCH-2179) Cleanup job for SOLR Performance Boost
[ https://issues.apache.org/jira/browse/NUTCH-2179?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15034511#comment-15034511 ] Sebastian Nagel commented on NUTCH-2179: +1: SolrIndexWriter should queue the deletions the same way as done for additions/updates. Looks like the bulk commit by an UpdateRequest is already assumed because numDeletes is taken into account when checking whether the batchSize is reached (SolrIndexWriter, line 125: {{if (inputDocs.size() + numDeletes >= batchSize) {}} > Cleanup job for SOLR Performance Boost > -- > > Key: NUTCH-2179 > URL: https://issues.apache.org/jira/browse/NUTCH-2179 > Project: Nutch > Issue Type: Improvement > Components: indexer >Affects Versions: 1.9, 1.10, 1.11 >Reporter: David Johnson >Priority: Minor > Labels: patch > Original Estimate: 2h > Remaining Estimate: 2h > > During a cleanup job, index deletes are scheduled one by one, which can make > a large job take days -- This message was sent by Atlassian JIRA (v6.3.4#6332)