[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16166020#comment-16166020 ] Steve Loughran commented on HADOOP-14303: - from stack overflow, evidence the aws xfer manager

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16015863#comment-16015863 ] Steve Loughran commented on HADOOP-14303: - see also HADOOP-11572 about how to handle exceptions

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16014601#comment-16014601 ] Steve Loughran commented on HADOOP-14303: - + S3aInputStream retries on some non-recoverable

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16014225#comment-16014225 ] Steve Loughran commented on HADOOP-14303: - Transient failure, seen on the s3guard+committer

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16006755#comment-16006755 ] Steve Loughran commented on HADOOP-14303: - I'm thinking if we target Java 8+, we could do this

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16006748#comment-16006748 ] Steve Loughran commented on HADOOP-14303: - HADOOP-13205 may be this too, though it's not as

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15989131#comment-15989131 ] Steve Loughran commented on HADOOP-14303: - +DELETE would seem something to look at here; you

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15973559#comment-15973559 ] Steve Loughran commented on HADOOP-14303: - +retry logic should include handling of any

[jira] [Commented] (HADOOP-14303) Review retry logic on all S3 SDK calls, implement where needed

[ https://issues.apache.org/jira/browse/HADOOP-14303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15967405#comment-15967405 ] Steve Loughran commented on HADOOP-14303: - If we could set the socket factory in the test JVMs,