You are viewing a plain text version of this content. The canonical link for it is here.
Posted to server-dev@james.apache.org by "Antoine Duprat (JIRA)" <se...@james.apache.org> on 2017/07/05 12:14:00 UTC
[jira] [Updated] (JAMES-2092) Pass Messages along with Added
mailbox events
[ https://issues.apache.org/jira/browse/JAMES-2092?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Antoine Duprat updated JAMES-2092:
----------------------------------
Component/s: (was: cassandra)
mailbox
> Pass Messages along with Added mailbox events
> ---------------------------------------------
>
> Key: JAMES-2092
> URL: https://issues.apache.org/jira/browse/JAMES-2092
> Project: James Server
> Issue Type: Bug
> Components: mailbox
> Reporter: Antoine Duprat
>
> Today, James performs a read just after write for retrieving a freshly added message in Cassandra, and index it in ElasticSearch.
> This behaviour is bad:
> - We generate unneeded latencies for Added operation (reading a message should be considered a long operation)
> - We generate read load on Cassandra on our write path
> - If the read fails and timeout, ElasticSearch is never updated. By having this read, we maximise chances for failure. As a reminder, if ElasticSearch index is not updated, the message is not visible in JMAP.
> We could pass along messages that were APPENDED, and thus save this costly, dangerous, error-prone write.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org