jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thomas Mueller (JIRA)" <j...@apache.org>
Subject [jira] Commented: (JCR-2396) File Journal workaround for misbehaving NFS implementations
Date Mon, 17 May 2010 07:58:43 GMT

    [ https://issues.apache.org/jira/browse/JCR-2396?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12868123#action_12868123

Thomas Mueller commented on JCR-2396:

> Was this patch applied to the 2.x line? 

No, the path was never applied in the trunk. The reason is: it turned out the patch didn't
completely solve the problem for the given NFS implementation. As far as I remember, the only
solution would have been to wait 2 seconds or more when reading zeroes. While it might be
possible to build a patch that also solves the remaining issues, I think that would be the
wrong solution in the long run. The code would get really complicated and hard to maintain.
Also, performance would be really bad (if the file system is not reliable). I think it's better
if we say the shared file system needs to be reliable.

I think the long term solution should be: don't rely on a shared file system except for very
simple operations. I guess a shared file system is not problematic for the file data store
and to bootstrap clustering (one properties file per cluster node that contains the IP address
and port).

> File Journal workaround for misbehaving NFS implementations
> -----------------------------------------------------------
>                 Key: JCR-2396
>                 URL: https://issues.apache.org/jira/browse/JCR-2396
>             Project: Jackrabbit Content Repository
>          Issue Type: Improvement
>          Components: clustering
>    Affects Versions: core 1.4.10
>         Environment: Some versions of NFS
>            Reporter: Thomas Mueller
>            Assignee: Thomas Mueller
>            Priority: Minor
>         Attachments: diff.patch
> After one NFS client wrote data to a file, reading from another client results in blocks
of zero
> in some NFS implementations. This is a known issue:
> http://markmail.org/search/blocks+of+zeros+%28NULLs%29+in+NFS+files+in+kernels
> The Jackrabbit file journal implementation does not work properly in this situation.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message