[Mageia-sysadm] Swap kills champagne this morning

Michael Scherer misc at zarb.org
Mon Mar 21 13:03:03 CET 2011


 On Sun, 20 Mar 2011 23:34:29 +0100, Damien Lallement wrote:
> On Sun, 20 Mar 2011 15:05:39 +0100, Michael Scherer <misc at zarb.org>
> wrote:
>> On Sat, 19 Mar 2011 20:30:29 +0100, Damien Lallement wrote:
>>
>>> FYI, all is now working as expected.
>>> There was a small issue with MySQL: a "mysqladmin flush-hosts" (on
>>> krampouezh) fixed it.
>>
>> You mean on champagne, or was the assignation of the various server
>> was changed ?
>
> No I was talking about krampouezh as it is the MySQL server as always
> said.

 That's not exactly what I understood :
 - there is no comment about using mysql on krampouezh in nodes.pp
 - it seems more logical to me to have the same server running sql and 
 apache
 for performance related reasons. ( ie, no network roundtrip )
 - it is also more logical to have less SPOF by having this on a single 
 system

 I may have missed mails on sysadm, but the only one I found date back 
 to
 october, saying that mysql will be used for blogs :
 https://www.mageia.org/pipermail/mageia-sysadm/2010-October/000121.html

 So sure, maybe you told to others people or on irc so you may see my 
 remark as unfair,
 but I think it would be nice to warn admins on the canonical 
 communication channel, ie
 here.

>> As krampouezh was planned to be used as a secondary mx, secondary
>> ldap and others stuff
>
> like MySQL server.

 See previous answer. And I think this should be avoided if possible as
 this mean we have 2 SPOF for the blogs, champagne and krampouezh.

 Given that krampouezh already stopped without notice ( due to
 a upgrade of a gandi related rpm, as you explained off-list ),
 I would prefer that we first find what exactly happened before 
 deploying
 a system where we can be hit 2 times more by the problem again. I 
 didn't
 upgraded neither krampouezh or champagne for that very reason :

 [root at krampouezh ~]# urpmq --auto-select
 gandi-hosting-agent
 gandi-hosting-vm
 libpostfix1
 postfix

 [root at web1 ~]# urpmq --auto-select
 gandi-hosting-agent
 gandi-hosting-vm
 libpostfix1
 postfix

 Of course, the problem is that champagne do not have enough ram, but 
 then I think
 we could rediscuss the usage of champagne since it doesn't fullfill 
 newer
 requirements in term of hosting.

-- 
 Michael Scherer


More information about the Mageia-sysadm mailing list