[ https://issues.apache.org/jira/browse/PHOENIX-3949?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
James Taylor updated PHOENIX-3949: ---------------------------------- Summary: Stop index maintenance until INDEX_DISABLED_TIMESTAMP is cleared (was: Stop index writing until index_disable_timestamp is cleared) > Stop index maintenance until INDEX_DISABLED_TIMESTAMP is cleared > ---------------------------------------------------------------- > > Key: PHOENIX-3949 > URL: https://issues.apache.org/jira/browse/PHOENIX-3949 > Project: Phoenix > Issue Type: Bug > Reporter: James Taylor > Labels: indexImprovement > > An index can be configured to remain online if a write failure occurs as our > partial index rebuilder will kick in upon failure and catch it up. However, > by keeping the index active, subsequent writes to the data table will likely > time out when they attempt to update the index given the current timeouts > (see PHOENIX-3948). > An alternative would be to hold off on any index updates until the partial > index rebuilder has completed the catch up process and cleared the > INDEX_DISABLED_TIMESTAMP field. This creates a race condition with writes > that happen right as the INDEX_DISABLED_TIMESTAMP field is cleared. We could > run a catchup query on the partial index rebuild after the field is cleared > to ensure we find rows that were committed before the regular index > maintenance starts again. -- This message was sent by Atlassian JIRA (v6.4.14#64029)