Migration Failed on upgrade from rc5 to stable [#2171571]

Hi,

As already posted yesterday in IRC i have a MEGA problem. I started a aegir migration from rc5 to stable. While running the migration i got a error message: https://gist.github.com/anonymous/a4a96c61344f018dc405 - After that my aegir was still running, but hosting queue failed and was still pointing to the old rc5. Also in the web frontend the complete configuration points still to rc5. While the webfrontend seems already to run under stable the configuration shows me that it's still rc5. I was able to run drush hostmaster commanfs after changing the path in .drush/hostmaster.alias.drushrc.php.
But i am not able to do a verify of the hostmaster site and i think the system is really broken.

What can i do to get a running aegir installation again?

Thank you very much!

Best regards

dob

Comment	File	Size	Author
#4	debug.txt	19.96 KB	dob_
	Log.txt	41.54 KB	dob_

Support from Acquia helps fund testing for Drupal Acquia logo

Comments

Comment #1

helmo CreditAttribution: helmo commented 12 January 2014 at 12:52

Note to others: I've already created #2171075: Fix hostmaster-resume command options after speaking with dob_ in IRC. The patch there might have prevented this situation.

Question now is how to recover.

Comment #2

helmo CreditAttribution: helmo commented 12 January 2014 at 13:06

Thanks for the full log.

Did you try the drush --debug @hostmaster hostmaster-resume platform_hostmaster6x20rc5 platform_hostmaster6x20 command?

or as ergonlogic suggested drush --debug @hostmaster hosting-task <nid> If you know the node id of a task...

Or drush --debug @hostmaster provision-verify

Please provide the full debug log.

Comment #3

dob_ CreditAttribution: dob_ commented 12 January 2014 at 14:23

Okay it seeems my hostmaster platform is missing.

I tried the following:
drush --debug @hostmaster hostmaster-resume platform_hostmaster6x20rc5 platform_hostmaster6x20
Added the debug.log to the issue.

Then i did:
drush --debug @hostmaster provision-verify

Got this error:
PHP Fatal error: Call to a member function get_services() on a non-object in /var/aegir/.drush/provision/Provision/Context.php on line 293

Fatal error: Call to a member function get_services() on a non-object in /var/aegir/.drush/provision/Provision/Context.php on line 293

Then my interface was gone. Vhost pointed to rc5 platform an site which is now empty.

Changed the platform_hostmaster.alias.inc to point to the stable and did a
drush --debug @hostmaster provision-verify

Then my interface was available again.

If i click in the platform section on hostmaster it points only to the url
http://aegir01.example.com/hosting/c/
and i get a page not found error.

Any ideas?

Comment #4

dob_ CreditAttribution: dob_ commented 12 January 2014 at 14:24

File	Size
debug.txt	19.96 KB

Comment #5

dob_ CreditAttribution: dob_ commented 14 January 2014 at 08:52

Solved the Problem reinstalling the server. ;-)

Aegir Docker installation

Comment #6

helmo CreditAttribution: helmo commented 14 January 2014 at 09:24

Status:

Active

» Fixed

I hope we can make this easier in the future ... atleast the root cause is fixed.

Comment #7

dob_ CreditAttribution: dob_ commented 14 January 2014 at 09:31

Helmo, i think the fix should be in the current stable version to prohibit that failure. Will you release a 2.0.1 as bugfix release?

Comment #8

28 January 2014 at 09:40

Status:

Fixed

» Closed (fixed)

Automatically closed - issue fixed for 2 weeks with no activity.

Comment #9

xurizaemon

he/him

Ōtepoti, Aotearoa 🏝

CreditAttribution: xurizaemon commented 9 March 2014 at 03:45

Status:

Closed (fixed)

» Needs work

Experienced the same issue. Some additional info ...

* In the web UI links for platform are to hosting/c, but the replacement platform could be found @ hosting/c/platform_hostmaster6x20
* {hosting_context} contained an entry with no name, causing the broken link to platform
* correcting the above (and removing one of two aliases for that node/%nid path in {url_alias} permitted the platform and hostmaster site to verify
* I may or may not have manually corrected a few other fields before doing the above

Theory: In Provision_Context::get_services(), if the hosting context doesn't have a name or name is empty, we end up with an empty $this->parent_key, are unable to reference the platform using $this->{$this->parent_key} and all go boom.

Moving back to needs work in case this wants a fix rather than a workaround. Still unsure how we ended up with a blank name in hosting_context, sorry!