accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "marco polo (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (ACCUMULO-4023) Table Ops appear to lock namespace and block other table ops
Date Thu, 08 Oct 2015 17:06:27 GMT

     [ https://issues.apache.org/jira/browse/ACCUMULO-4023?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

marco polo updated ACCUMULO-4023:
---------------------------------
    Description: I've noticed table repeatable operations impacting other table ops. Upon
looking at the code it appears as though we reserve the name space, attempt to lock it, and
then lock the table. Someone more familiar with the code can debunk me, but it appears as
though we're blocking table ops on alternate tables. The same code for getting the lock is
used for namespace and tables, it seems as though in most cases we only want a table lock,
we could change the path to be ztable_lock/namespaceId/tableId.   (was: I've noticed table
repo operations impacting other table ops. Upon looking at the code it appears as though we
reserve the name space, attempt to lock it, and then lock the table. Someone more familiar
with the code can debunk me, but it appears as though we're blocking table ops on alternate
tables. The same code for getting the lock is used for namespace and tables, it seems as though
in most cases we only want a table lock, we could change the path to be ztable_lock/namespaceId/tableId.
)

> Table Ops appear to lock namespace and block other table ops
> ------------------------------------------------------------
>
>                 Key: ACCUMULO-4023
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-4023
>             Project: Accumulo
>          Issue Type: Improvement
>    Affects Versions: 1.6.4
>            Reporter: marco polo
>
> I've noticed table repeatable operations impacting other table ops. Upon looking at the
code it appears as though we reserve the name space, attempt to lock it, and then lock the
table. Someone more familiar with the code can debunk me, but it appears as though we're blocking
table ops on alternate tables. The same code for getting the lock is used for namespace and
tables, it seems as though in most cases we only want a table lock, we could change the path
to be ztable_lock/namespaceId/tableId. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message