update process

[danliker]

why does each plesk update takes hour's for the process ?

 

[IgorG]

Usually it depends on quantity of mailboxes and domains.

 

[danliker]

that's right, but how can we reduce the dowtime, it is horrible!

what do big companys with many domains/mailboxes on each server ?

 

[IgorG]

Unfortunately mailsettings rebuilding and synchronization is absolutely necessary operation and it can't be disabled or something else. Therefore I can recommend you increase power of server's hardware, disabling and removing all unused domains/mailboxes, migrate some domains with mailboxes to other Plesk servers for decreasing quantity of mailboxes on server which should be upgraded and schedule upgrade downtime to night hours with corresponding customer's notification.

 

[danliker]

ok, but why is apache stopped during rebuilding and synchronization of mailboxes ?

 

[IgorG]

How apache related packages can be upgraded without stopping httpd processes? And all vhosts configs should be rebuilded by websrvmng too.

 

[danliker]

don't you like to understand ?

during the prozess of mail update and rebuilding and synchronization of mailboxes, why must apache be stopped in this moment ?

i know that apache is restartet and websrvmng is running, but not on the SAME time !

as i see it is as fallow...

- do some updates
- stop apache
- update mailsystem
- rebuilding and synchronization of mailboxes
- update apache things
- start apache

i think id sould be:

- do some updates
- update mailsystem
- rebuilding and synchronization of mailboxes
- stop apache
- update apache things
- start apache

 

[IgorG]

I can assume that it is related to webmails because webmails have config dependences for mailsettings and for websettings too.

 

[surfichris]

Respectfully I disagree. The Plesk update system is poorly structured if you ask me.

Why does every single update need to rebuild the mail configuration and DNS configuration? I'm almost certain that EVERY release does not need to do this. It should be done where necessary. This is easy enough to do - you just set some DB flags or flags in a file then when processing the packages and running the post-install you do what you need to.

Another issue I have is the fact that Plesk rebuilds this information SEVERAL times:

..
(Reading database ... 54566 files and directories currently installed.)
Preparing to replace plesk-base 9.2.3-debian4.0.build92091009.18 (using .../plesk-base_9.3.0-debian4.0.build93091230.06_amd64.deb) ...
...
Trying to stop service qmail... done
Unpacking replacement psa-qmail ...
Preparing to replace psa-qmail-rblsmtpd 0.70-debian4.0.build92091009.18 (using .../psa-qmail-rblsmtpd_0.70-debian4.0.build93091230.06_amd64.deb) ...
Unpacking replacement psa-qmail-rblsmtpd ...
Setting up psa-courier-imap (3.0.8-debian4.0.build93091230.06) ...
Trying to register service xinetd... using /usr/sbin/update-rc.d
System startup links for /etc/init.d/xinetd already exist.
done
Trying to executing mail_restore to synchronize mail server settings and Plesk Database...
done

Setting up psa-mail-driver-common (9.3.0-debian4.0.build93091230.06) ...
mv: cannot stat `/var/qmail/control/spf*': No such file or directory
Trying to executing mail_restore to synchronize mail server settings and Plesk Database... done
...
Setting up psa-qmail (1.03-debian4.0.build93091230.06) ...
...
Trying to start service qmail... done
...
Trying to set psa database version to 0930... Trying to stop service bind9... Stopping domain name service...: bind
named: no process killed
named: no process killed
failed!
done
Plesk rebuilds all DNS zones here
Trying to start service bind9... done
Trying to clean up compiled python modules... Trying to update Components table... done
Trying to executing mail_restore to synchronize mail server settings and Plesk Database... done

Trying to start Plesk...
Stopping Plesk... done

what do big companys with many domains/mailboxes on each server ?

We restart Apache as soon as we can and don't give a damn what Plesk is doing. We normally restart it as soon as we notice it's down and do so by running the websrvmng -a -v command to make sure Apache has the correct configuration.

Therefore I can recommend you increase power of server's hardware

So in my dual processor, quad core Xeon servers with 24GB/32GB RAM and RAID 10 over SAS disks, what specs should I look at moving to? Because we don't run your "typical" hosting company boxes.

Surely that's not a recommendation. :-/

How apache related packages can be upgraded without stopping httpd processes?

How can't they? Replace binaries, rebuild config and then restart. The beauty of Linux is that you can replace an open/running file without any adverse affects.