infra-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sebb (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (INFRA-16342) mail-archives: missing mbox and index files
Date Tue, 24 Apr 2018 09:32:00 GMT

    [ https://issues.apache.org/jira/browse/INFRA-16342?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16449525#comment-16449525
] 

Sebb commented on INFRA-16342:
------------------------------

I think graduations and renames can probably still cause the problem on mail-archives, because
that only processes the current month.

AFAIK at present a list rename is not an atomic operation as far as the mino archives are
concerned.
In any case, the older archives have earlier dates so will not get indexed on mail-archives
when they are synched across from the new location. The rename of the existing archives is
a later manual operation currently.

If a cronjob ever fails to complete there might be further un-indexed files if it is around
month end when the archives are compressed on minotaur.

The problem is that mod-mbox-util -u only checks if the mbox file is newer than listinfo.db.
It does not check the msgsum index for the mbox, which may be old or missing.

The existing mail-archives scripts don't take this into account.
The mail-private scripts are better in that respect.

> mail-archives: missing mbox and index files
> -------------------------------------------
>
>                 Key: INFRA-16342
>                 URL: https://issues.apache.org/jira/browse/INFRA-16342
>             Project: Infrastructure
>          Issue Type: Bug
>          Components: Mail Archives
>            Reporter: Sebb
>            Priority: Major
>
> As part of checking whether it was OK to delete graduated and renamed incubator mbox
files, I found several TLP directories which did not have a full complement of mbox files.
Many of the early gz files did not have a corresponding uncompressed file nor an mbox link
to it.
> The following directories were affected:
> vcl-*
> vxquery-*
> wink-*
> wookie-*
> The missing files and links have been created and the indexes recreated.
> The cause is probably that the jobs that unpack the files and update the indexes only
process the current (or previous) month. If a compressed file for another month is added to
minotaur (e.g. due to graduation or rename) it won't be processed.



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

Mime
View raw message