[Mailman-Users] Attachments are unexpectedly re-created.

Mark Sapiro mark at msapiro.net
Tue Mar 2 22:22:21 CET 2010

Masaharu Kawada wrote:
>Firtst of all, one of my customer wanted to delete attachments to
>make more disk spaces on the server. The server's disk resource
>was getting full because the size of the directory for attachments
>was very big. So the customer did the following works first.
>Digest options -> Disabled(Digestable 'NO')
>Archiving options -> Disabled(Archive 'NO')
>Secondly, after the above, the customer deleted directories located
>under the following directory such as 20091201, 20091202 and so on.
>/var/lib/mailman/archives/private/<list name>/attachments/<date>
># rm -rf 20091201,20091202,.....
>Finaly, the customer got the "Digest options" and "Archiving options"
>back to enabled as follows.
>Digest options -> Enabled(Digestable 'YES')
>Archiving options -> Enabled(Archive 'YES')

There was no need to turn off digestable and archive if they were only
going to be turned on again.

>With this operations, the directories under attachments directory
>were deleted and the disk space is now enough, however, by unknown
>timing, directories that are named with past's date, which were
>manually deleted, are re-created automatically. And in that 
>directories, it seems that the same attachment files are repeatly
>One thing that I realized to stop this weird action is to disable
>both "Digest options" and "Archiving options". Once those options
>are disabled, no more unexpected attachment files would be created
>under that directory. 
>To know what was going on to the mailman server at that moment
>(in the meantime of the unexpected action), I asked the customer
>for logs such as error log in /var/log/mailman direcoty, and 
>in the error log, I saw the following errors regarding "send_digests".
>xxx xx xx:xx:xx 2010 (2129) send_digests() failed: EUC-JP decoding error:
>invalid character ......
>xxx xx xx:xx:xx 2010 (2129) send_digests() failed: EUC-JP decoding error:
>invalid character ......
>xxx xx xx:xx:xx 2010 (2129) send_digests() failed: EUC-JP decoding error:
>invalid character ......
>I have not yet been sure if these errors are related to the cuase,
>but it seems that "send_digests()" appearently fails and retries

Yes, I think this is the problem. Somewhere in the
lists/LISTNAME/digest.mbox file there is a bad message that causes
this error. Every time there is a post to the list, A digest is
triggered because of the size of the digest.mbox. The mbox is
processed and attachments are scrubbed (from the 'plain' digest) and
stored up to the point of the exception which aborts the process until
the next time, but each time those attachments prior to the bad
message get stored again.

>My questions this time are as follows.
>1.Why are attachment files re-created? Is it a normal action whenever
>  "Digest options" and "Archiving options" are available?

Non text/plain MIME parts and text/plain MIME parts with unknown
character set are removed and stored aside and replaced by links in
both the archive and the plain format digest. Thus there normally are
two copies of each in the archive.

In your case, the same messages in the digest.mbox are processed
repeatedly causing the attachments to accumulate.

>2.If the answer for the question 1 is "NO", what could be the cause?
>  The fail of send_digests() is it?
>3.Does the send_digests() repeat(retry) if the action fails?
>4.Is there anything else that should be done after deleting attachments
>   manually to avoid being suffered this weird behavior?

The underlying cause is a message in the lists/LISTNAME/digest.mbox
file with a body or sub part with declared as charset=EUC-JP with an
invalid character or something related.

Removing or moving aside the lists/LISTNAME/digest.mbox or finding the
bad message and removing it from the file should also fix this.

I consider this behavior to be a bug. I'll look into fixing it. I think
what I'll need to do is just move the digest.mbox aside and log an
error message. It would be better if I could just bypass the bad
message, but that may be difficult.

Mark Sapiro <mark at msapiro.net>        The highway is for gamblers,
San Francisco Bay Area, California    better use your sense - B. Dylan

More information about the Mailman-Users mailing list