[Mageia-sysadm] Disk full on valstar

Mon Jan 24 14:02:51 CET 2011

On Mon, Jan 24, 2011 at 12:41, Olivier Blin <mageia at blino.org> wrote:
+> Pascal Terjan <pterjan at gmail.com> writes:
+>
+>> On Mon, Jan 24, 2011 at 07:39, Olivier Blin <mageia at blino.org> wrote:
+>>> nicolas vigier <boklm at mars-attacks.org> writes:
+>>>
+>>>>> > - do not keep src.rpm in case of failure. We do not really use them
+>>>>> > anyway, and removing them freed 6g on the server.
+>>>>>
+>>>>> We could have a cron, like on Mandriva BS, removing files older than a
+>>>>> few days (like 7 days).
+>>>>
+>>>> Or one month, or more (if it can be useful). We have 1.53T free so we
+>>>> can probably create a big lvm partition for this.
+>>>>
+>>>> Maybe we can create a 100G lvm partition for this ?
+>>>
+>>> It grew to about 14G in 2 weeks, without big packages like OpenOffice or
+>>> OpenArena, we have to be careful.
+>>>
+>>> Also, having more src.rpm/.info files in /uploads means the BS and its
+>>> web status will be slower.
+>>>
+>>> Maybe we can move them to a /olds-uploads dir after one week, to make
+>>> the BS faster but still keep history and remove them from there after
+>>> one moth.
+>>
+>> Why not keeping at most n submits (like 10K) and at most 1 month and
+>> at most 90G ?
+>> If we get over 90G remove oldest ones, day by day, until we get back
+>> into the limit
+>
+> Maybe we could use this, but with two levels (so that uploads + web
+> interface are not slowed down too much):
+> - in /uploads, at most 500 submits or 1 week or 20G
+> - in /old-uploads, at most 10k submits or 1 month or 90G
+
+Well there is no real reason to keep any uploaded or failures things
+in upload anyway
+