Re: Cyrus Murder Environment Upgrade

[Miguel Mucio Santos Moreira]

Wolfgang,

I'm sure your help and experience with Cyrus Murder upgrading will save a lot of time and reduce the possibility of an eventual problem during the upgrade.

Thankful

--

Miguel Moreira
DTE/SRE/GRE - Gerência de Redes
+55(31)3339-1401
PRODEMGE - Companhia de Tecnologia da Informação do Estado de Minas Gerais


Aviso: Esta mensagem é destinada exclusivamente para a(s) pessoa(s) a quem é dirigida, podendo conter informação sigilosa e legalmente protegida. O uso impróprio será tratado conforme as normas da empresa e a legislação em vigor. Caso não seja o destinatário, favor notificar o remetente, ficando proibidas a utilização, divulgação, cópia e distribuição. Em Terça, Junho 09, 2020 10:39 -03, Wolfgang Breyha <wbreyha at gmx.net> escreveu:Hi!

On 09/06/2020 14:56, Miguel Mucio Santos Moreira wrote:
> Dear Wolfgang,
>
> Firstly thanks for your answer, secondly I have one more doubt, during this
> time where the new Mupdate Master is receiving mailboxes information from
> backend servers, is necessary stopping comunications between frontend
> servers and mupdate master or none action is necessary besides that one
> you've mentioned before?
> We're concerned if frontend servers will connect to Mupdate Master and
> receive from it an information which there's no mailboxes anymore until the
> backend push entirely mailboxes information and this situation to cause any
> trouble.

I recommend the follow steps:
*) check that your currently running setup has no conflicts in mailboxes.db
by running "ctl_mboxlist -mw". It is possible that you see output if
somebody changes his mailboxes while you run the command, but it should
not appear again if you run the command a second time. If everything is
fine...

*) shut down all running cyrus
frontends first, then backends and mupdate last
*) backup your mailboxes.db on mupdate server
*) replace/update mupdate and start it with empty/removed mailboxes.db.
*) backup your mailboxes.db on the backends
*) do the ctl_mboxlist -m by hand on the backends (only one at the same
time) you can check if everything went ok with "-mw" at any time
afterwards.
*) start cyrus on backends

In our setup the frontends have mupdate running as well. I can't currently
remember if this is mandatory. If this is true in your setup then:
*) remove or rename the mailboxes.db database on the frontends and start
them. They will fetch the database immediately from the mupdate server.
This is visible in syslog as well and takes about 30 seconds in our
setup (~250MB mailboxes.db).
otherwise
*) start the frontends

At this point everything should be up and running again.

Watching syslog output all the time usually helps.

IIRC we updated the mupdate server to 2.5 first. Then we had a mix of 2.4
and 2.5 backends for some time. The 2.5 backends had some capabilities
suppressed. Frontends had 2.4 until all backends had 2.5. Last but not
least we upgraded to 2.5 on the frontends.

If you need more info/details feel free to ask.

Greetings, Wolfgang
--
Wolfgang Breyha <wbreyha at gmx.net> | https://www.blafasel.at/
Vienna University Computer Center | Austria
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/info-cyrus/attachments/20200609/e4ac14e2/attachment-0001.html>