Savannah Administration - News: lists.gnu.org delays
Item posted by Sylvain Beucler <Beuc> on Wed 01 Mar 2006 07:42:44 AM UTC.
lists.gnu.org is experiencing issues to due a database corruption.
Check the comments for more information.
Also, lists.gnu.org is being revamped. ETA is fall of March. The Savannah Hackers are not directly involved in that process. We'll post additional information here as available. The system will improve on delays and spamming.
2006-02-25: original post
2006-03-01: update and precisions
Messages lost? (posted by Eric NOULARD, Sun 19 Mar 2006 10:18:47 PM UTC) |
I think we are facing message loss recently
on tsp-devel@nongnu.org.
Some messages which were directly sent to the list
and CC to some other people did reach the people
and not the list.
Tonight I did send 2 messages to the list
the second reached the list but the first did not
seem to go through?
May those trouble be related to the lists.nongnu.org
recent ones?
|
Update (posted by Sylvain Beucler, Wed 01 Mar 2006 07:32:32 AM UTC) |
Here are some informations about what is going on.
On 23 february, one of the database files for the tmda system on lists got corrupted. As a result, any attempts to deliver messages to tmda failed, and the mail server simply started queuing them. This queue started piling up pretty quickly; the mail server software handled it - until the mail log exceeded the limit of 2GB, from all of the retries to deliver all of the queued messages to tmda. And so the mail server simply quit delivering mail.
Jag moved the 2GB mail log out of the way and reinitialized the tmda database and is now figuring out what exactly to do with the 1.6 million messages in queue. First he'll prune them down for obvious spam -- paypal adverts and the like -- hopefully getting that down to about 900k before i start delivering again. Then there will be some delays again until the system return to normal operation.
Some additional temporary measures to be more aggressive about spam are likely to be setup this week, maybe greylisting (the inherent additional delay will not be really noticeable in this particular case).
In this light TMDA will not be a feature of the under-construction new mail system.
Feel free to add comments here, we'll try to sum them up.
|
RE: Update (posted by Sylvain Beucler, Wed 01 Mar 2006 07:38:47 AM UTC) |
We saw (and were reported about) duplicate deliveries. We notified lists.gnu.org admins.
|
RE: Update (posted by Sylvain Beucler, Thu 02 Mar 2006 08:56:06 PM UTC) |
Some information from Justin (that was yesterday):
Right now we're working on eliminating the duplicate messages from the queue, but this will take a while, and requires us having sole access to the queue itself, so unfortunately we're having to defer list mail at the moment while we fix the queue.
When TMDA broke, due to the sheer volume of tmda messages per day that are processed (somewhere on the order of 50k) and the fact that the router was deferring them over and over, the logs filled... After this point, Exim seems to have still queued messages but didn't actually return a successful delivery code to another machine in the mail system - thus, the duplicates.
|
RE: Update (posted by Sylvain Beucler, Wed 08 Mar 2006 11:35:04 PM UTC) |
List archives are updating again.
|
RE: Update (posted by Simon Waters, Sat 04 Mar 2006 10:36:31 AM UTC) |
I'm doing greylisting, and using the sbl-xbl blacklist, and outperforming the old list server at sorting spam by a very obvious margin.
Appreciate these things are hard to scale, but as an approach it is both simple, and avoids the need for content inspection.
Thanks for the update, wondered why some list messages were 24 hours old, and keep up the good work.
|
what about pending emails? (posted by Claudio Fontana, Mon 27 Feb 2006 01:01:44 PM UTC) |
Hello,
I have many pending emails to projects' mailing lists,
which are very important to me.
Could the pending mails be delivered, please?
Thanks
CLaudio
|