[ 
https://issues.apache.org/jira/browse/CONNECTORS-1600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16824738#comment-16824738
 ] 

Tang Huan Song commented on CONNECTORS-1600:
--------------------------------------------

Setting retriesRemaining=5 and totalTries=8 enabled the job to complete in a 
reliable fashion. It is likely that the heavy traffic subsided in the 
additional retries, allowing for the requests to succeed.

However, I acknowledge that luck played a part in the job's successes, and I am 
of the opinion that changing the value doesn't alleviate the high-traffic 
issue, but merely works around it. Though it would be beneficial to us were the 
retry counts made configurable, I have no objections if you choose not to 
implement this.

Thanks for your thoughts!

> Add support for configuring JCIFS connector's resilience to SMB exceptions 
> before throwing a ServiceInterruption
> ----------------------------------------------------------------------------------------------------------------
>
>                 Key: CONNECTORS-1600
>                 URL: https://issues.apache.org/jira/browse/CONNECTORS-1600
>             Project: ManifoldCF
>          Issue Type: Improvement
>          Components: JCIFS connector
>    Affects Versions: ManifoldCF 2.10, ManifoldCF 2.11, ManifoldCF 2.12
>            Reporter: Tang Huan Song
>            Priority: Major
>
> This is a improvement request regarding the JCIFS(-ng) connector's exception 
> handling behavior.
> After examining the JCIFS connector code, I've found that the number of 
> retries given consecutive identical SMB exceptions and the total number of 
> retries per file/request is hardcoded within the connector at 
> retriesRemaining=3 and totalTries=5 respectively.
> Depending on the amount of traffic a file server regularly handles, the 
> probability of any given SMB request failing, and correspondingly the total 
> number of SMB request failures for a given file request will vary. As a 
> result, the current hardcoded values may cause ManifoldCF to abandon the job 
> in the event of high traffic.
> I would like to suggest making these values configurable, as a connector-wide 
> setting modified via ManifoldCF's properties.xml or a per-connection setting 
> modified via the corresponding repository connection's page.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to