www-repository mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brian E. Fox" <bri...@reply.infinity.nu>
Subject RE: Turbine artefacts not mirrored
Date Thu, 11 Dec 2008 23:30:58 GMT

>The discussion is in the Maven PMC private list, so I wouldn't like to
>give details on why/how the sync stopped, hoping that the people
>involved comment here in the future.

Nothing sinister, just an oversight.

Basically the repo machine was out of disk and because of the heavy
traffic we didn't want to take it out to install the new disk without
some failover in place first. It was taking a while to get some
balancing and failover in place so we looked at ways to reduce the disk
usage as a stop-gap.

The M1 repo was actually being served via rewrite rules that pull from
the M2 data, yet the repo was sitting there taking up a huge amount of
disk. So it was decided to take out the M1 repo, which also stopped the
sync. When I made the switch to Nginx, I wasn't aware of the rewrite
rules and it was more than four days before anyone noticed that the m1
repo was broken. This lead us to believe that not many people were using
M1 and we certainly didn't expect that new artifacts were being deployed
via the M1 repo.

Rather than sync in the entire apache m1 repo for nothing, we should
handle the remaining m1 deploys on a case by case basis and try to get
them into the m2 format since that will serve both m2 users and m1 via
the rewrites. Several suggestions have been kicked around about how that
can be done. I'm not super familiar with the conversion process that was
in place so I won't comment on the actual mechanics.


View raw message