Savannah Administration - News
Savannah status
Item posted by Sylvain Beucler <Beuc> on Thu 11 Jun 2009 05:37:06 PM UTC.
All services are running fine now, except for a few things to fix in webcvs replication.
CVS/SVN/Git/Hg: no need to send new recovery requests, all repos are restored, either from your instructions, either from the May 27 backup.
Arch/Bzr: restored from May 29
Downloads: restored from May 29
WebCVS; restored from May 27 at Savannah, refreshed accordingly at www.gnu.org and www.nongnu.org. For www.gnu.org (group 'www') only, changes between 27 and 29 were manually re-committed. If you need a diff of these changes for your project, please send a support request.
The May 27 backup was found some days after the crash because of ongoing changes in the FSF backup system. The location of the latest backups are now properly documented.
The origin of the crash is still unknown. A first degradation was due to a broken disk. The one that happened the next day and corrupted the whole disk array is unknown. The experts we sent the disks to could not easily recover a whole directory structure, so we will not further attempt to recover data from the crashed disks.
Currently we're planning a move to newer hardware to avoid taking any risk. We'll also put an emphasis on having backups available in the same facility, to reduce the delay we got during the recovery. We'll also setup a DRBD real-time synchro to minimize the amount of data lost between the last backup and a crash.
Powered by Savane 3.14-5bfe.
Corresponding source code