FC4 -> FC5 Upgrade + Plesk

consultant · Unread post by **consultant** » Tue Mar 13, 2007 6:16 am

Hi Scott,

First of all I wanted to say thanks for a fantastic rpm archive that made my life easy a couple of days ago. I followed this tutorial:

http://elliottback.com/wp/archives/2007 ... d-plesk-8/

Which seems to have worked really well (apart from a few dependency issues which I forced).

The only thing that got trashed was Horde/IMP so webmail is completely hosed. After looking at the logs and errors I get, these are the highlights.

On first login attempt, after successful login I get bounced to an empty page (proper empty, not a single word of html). If I try to go back to the login page, it recognizes that I'm authenticated and gets stuck in an infinite redirect loop, which eventually gives up and dies, with the standard The browser redirection can never complete" etc

The web server log shows the following:

Code: Select all

[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP Fatal error:  Allowed memory size of 8388608 bytes exhausted (tried to allocate 368640 bytes) in /etc/psa-horde/horde/prefs.php on line 545, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP Stack trace:, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   1. {main}() /usr/share/psa-horde/imp/redirect.php:0, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   2. IMP_Session::createSession() /usr/share/psa-horde/imp/redirect.php:150, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   3. Auth_imp->authenticate() /usr/share/psa-horde/imp/lib/Session.php:224, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   4. Auth->authenticate() /usr/share/psa-horde/imp/lib/Auth/imp.php:120, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   5. Auth->setAuth() /usr/share/psa-horde/lib/Horde/Auth.php:182, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   6. Auth->clearAuth() /usr/share/psa-horde/lib/Horde/Auth.php:687, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   7. Registry->unloadPrefs() /usr/share/psa-horde/lib/Horde/Auth.php:786, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   8. Registry->loadPrefs() /usr/share/psa-horde/lib/Horde/Registry.php:986, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP   9. Prefs_session->retrieve() /usr/share/psa-horde/lib/Horde/Registry.php:944, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP  10. Prefs->retrieve() /usr/share/psa-horde/lib/Horde/Prefs/session.php:49, referer: http://webmail.example.co.uk/imp/login.php
[Mon Mar 12 23:58:00 2007] [error] [client xxx.xx.xxx.x] PHP  11. Prefs->_setDefaults() /usr/share/psa-horde/lib/Horde/Prefs.php:538, referer: http://webmail.example.co.uk/imp/login.php

If anyone has any ideas about what might be happening I'd be most appreciative.

Cheers

T.

breun · Unread post by **breun** » Tue Mar 13, 2007 7:03 am

I'd say the first line says it all. Raise your memory_limit for PHP (to 32M or so). Plesk for FC5 has a psa-php5-configurator package that sets your PHP settings, but maybe you don't have that installed as you previously had PHP4 and psa-php4-configurator installed (and the yum upgrade didn't fix this)?

psa-php4-configurator sets the following in your /etc/php.ini:

- sets memory_limit = 32M
- adds '.' to include_path
- sets short_open_tag = On
- sets file_uploads = On
- sets safe_mode = On

I don't know if psa-php5-configurator does anything different. You should just be able to install the rpm and have that set Plesk's defaults.

Unread post by **scott** » Tue Mar 13, 2007 8:13 am

yeah I'll bet it updated the php.ini and you've got an an .rpmnew one sitting there waiting to be mv'd.

consultant · Unread post by **consultant** » Wed Mar 21, 2007 6:39 am

scott wrote:yeah I'll bet it updated the php.ini and you've got an an .rpmnew one sitting there waiting to be mv'd.

You were right of course

Thanks for the pointers guys!! Just gotta figure out how to do RBLs now

I come from the world of postfix (awesome product, and currently maintainted :p ) So this qmail stuff is a little odd.

Cheers

T.

breun · Unread post by **breun** » Wed Mar 21, 2007 7:04 am

You can add RBL's in the Plesk web interface or use them in your SpamAssassin config.

jdaustin · Unread post by **jdaustin** » Wed Mar 28, 2007 9:00 pm

Anyone tried this on a 1and1 server?
The last time I messed with a kernel it hosed the network due to missing module.

Highland · Unread post by **Highland** » Thu Mar 29, 2007 9:20 am

I wouldn't recommend it on a 1and1 server running a 1and1 image. Their FC4 images aren't standard installs (do a df -h and you can see one of their "ideas"). The only way off FC4 is to reimage to Suse or Debian or to use Scott's CentOS 4 installer (which doesn't work on Business and Enterprise servers, only Root).

jdaustin · Unread post by **jdaustin** » Fri Mar 30, 2007 12:33 am

I hate being stuck on FC4 knowing that it's no longer supported.
I suppose I could compile everything from source to keep things more secure but then I'd be in dependancy hell with things that are still maintained.
I wish they'd had the option of a Centos image.
I can't beat or meet the $50/month I'm leasing this server for so I'm kind of stuck

consultant · Unread post by **consultant** » Fri Mar 30, 2007 4:31 am

Highland wrote:I wouldn't recommend it on a 1and1 server running a 1and1 image. Their FC4 images aren't standard installs (do a df -h and you can see one of their "ideas"). The only way off FC4 is to reimage to Suse or Debian or to use Scott's CentOS 4 installer (which doesn't work on Business and Enterprise servers, only Root).

I have to disagree with Highland, the upgrade I performed that I originally spoke about was on a 1 and 1 box, I made sure that it was clean and fresh, and performed it to the virgin equipment (i.e. Virgin install of the 1&1 platform)

I followed the link I pasted into my first post, and used Scotts Yum archives to take care of the plesk. Apart from a few troubles of my own causing and not reading the docs properly, it was surprisingly smooth. I'm not upgrading to FC6 because Plesk doesn't support it yet.

The only thing I will say is use the kernel they give you (i.e. the FC4 2.6.17 kernel) because they compile stuff in that isn't in the standard FC kernel. That said they have an FAQ online as to how to build a kernel for a 1&1 box (except it doesn't work) by using their config file which obviously lives in /boot/. But don't despair, on my previous 1&1 box I regularly compiled my own kernel successfully many times without incident. It took me a few goes to get it right, but it's only a big PC fer gawd sake, it's not undoable.

Since they got rid of the serial console access it's harder to debug why a kernel fails, and it's a lot more farting around to get the system back than simply picking a different kernel to boot from at boot. But for this the lilo boot log is your friend.

Just my 2 cents/1 pence

T.

breun · Unread post by **breun** » Fri Mar 30, 2007 4:40 am

consultant wrote:I'm not upgrading to FC6 because Plesk doesn't support it yet.

Plesk 8.1.1 was released a couple of days ago and it supports FC6.

consultant · Unread post by **consultant** » Fri Mar 30, 2007 6:45 am

breun wrote: Plesk 8.1.1 was released a couple of days ago and it supports FC6.

Cool!! Thanks for the heads up on that.... is there a what's new changelog anywhere that you know of?? I've had a little look, but can't find one yet.

Cheers

T.

consultant · Unread post by **consultant** » Fri Mar 30, 2007 6:51 am

Forget that, found it

Ta.

breun · Unread post by **breun** » Fri Mar 30, 2007 7:05 am

Release notes, autoinstaller and rpm .tar.gz archives can be found at http://www.swsoft.com/en/download/plesk8/ as usual.

Highland · Unread post by **Highland** » Fri Mar 30, 2007 9:19 am

I'm very impressed, consultant. I really didn't think it would work just because of all their non-stock stuff (the FC4 images didn't even work when they first rolled them out). You might want to post a walkthrough over at 1and1faq.com because people ask all the time how to do it and I advise reimaging rather than upgrading.

Since they got rid of the serial console access

??? Mine still works.

consultant · Unread post by **consultant** » Fri Mar 30, 2007 9:35 am

??? Mine still works.

Don't be impressed with me, be impressed with the dude that owns the link I pasted in my first post, he's the smart one. Also it was on a brand new box. I'd be loathe to try FC5 to FC6 now that the equipment is production live. The saftey net obviously with a brand new server is that I could take risks with forcing deps (which I only needed to do once) and if it went pear shaped I could simply re-image.

Regarding the remote console, I bought this server a few weeks ago, and they have a rescue image they use (I'm guessing using PXE boot or something) but I can find no links in my account page to a serial console any more. This means if I boot a bad kernel, firstly I can't see if it's a bootloader/initrd problem or if it's a failed driver or something else. Secondly I have to remote reboot, and use a new image to boot to (can take 10 mins to come back) login, mount original root on /mnt; modify lilo.conf; chroot; run lilo; then issue a remote reboot command again (takes up to 10 mins again) if you don't use their control panel it forces a net boot again

Cheers

T.