[ https://issues.apache.org/jira/browse/KUDU-2179?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Andrew Wong updated KUDU-2179: ------------------------------ Description: When ksck runs, it selects a single timestamp and does a snapshot scan at this time across all tablets. If the scans run for a long time (e.g. due to heavy traffic to the tservers), some scans may be attempted on data that has already been GC'ed, surfacing the errors: {{Error: Invalid argument: Snapshot timestamp is earlier than the ancient history mark. Consider increasing the value of the configuration parameter --tablet_history_max_age_sec. Snapshot timestamp: P: 1507232752670708 usec, L: 0 Ancient History Mark: P: 1507232752970869 usec, L: 0 Physical time difference: -0.300s}} This could be remediated by batching these scans and selecting a new timestamp for each batch. was: When ksck runs, it selects a single timestamp and does a snapshot scan at this time across all tablets. If the scans run for a long time (e.g. due to heavy traffic to the tservers), some scans may be attempted on data that has already been GC'ed, surfacing the errors: {{ Error: Invalid argument: Snapshot timestamp is earlier than the ancient history mark. Consider increasing the value of the configuration parameter --tablet_history_max_age_sec. Snapshot timestamp: P: 1507232752670708 usec, L: 0 Ancient History Mark: P: 1507232752970869 usec, L: 0 Physical time difference: -0.300s}} This could be remediated by batching these scans and selecting a new timestamp for each batch. > Have ksck not use a single snapshot for all tablets > --------------------------------------------------- > > Key: KUDU-2179 > URL: https://issues.apache.org/jira/browse/KUDU-2179 > Project: Kudu > Issue Type: Improvement > Components: ksck > Reporter: Andrew Wong > > When ksck runs, it selects a single timestamp and does a snapshot scan at > this time across all tablets. If the scans run for a long time (e.g. due to > heavy traffic to the tservers), some scans may be attempted on data that has > already been GC'ed, surfacing the errors: > {{Error: Invalid argument: Snapshot timestamp is earlier than the ancient > history mark. Consider increasing the value of the configuration parameter > --tablet_history_max_age_sec. Snapshot timestamp: P: 1507232752670708 usec, > L: 0 Ancient History Mark: P: 1507232752970869 usec, L: 0 Physical time > difference: -0.300s}} > This could be remediated by batching these scans and selecting a new > timestamp for each batch. -- This message was sent by Atlassian JIRA (v6.4.14#64029)