Issue [12.5.30] Plesk Panel migrator loops on a domain (because of horde database missing?)

[burnley]

Plesk 12.5.30 on CentOS on both source and target server, using migrator 2.7.9. I see the latest is 2.8.7 so I might have to look into it later down the track.
I think we have an issue which has been plaguing us recently. Some domains fail to migrate, causing the deployer process to loop for a long time, read *hours* before bailing out. Looking into the migration rsession logs I've found this error interesting (personal data obfuscated):

[2017-06-14 08:58:32.331|19576] INFO: Request is ready for: PMM_PROFILE_LOG=/usr/local/psa/PMM/rsessions/20170614074441590/profile.log PLESK_RESTORE_MODE=1 PLESK_DISABLE_PROVISIONING= LANG=en_US.UTF-8 ALLOW_WEAK_PASSWORDS= /usr/local/psa/admin/plib/api-cli/turboaddr.php '-c' '69c30770aa10566bb73c6f067c43b2c4' '-owner-email' '[email protected]' '-type' 'object' '-home-phone' '' '-fax' '' '-title' '' '-mobile-phone' '0123 456 789' '-company' 'Company Name' '-email' '[email protected]' '-alias' '' '-notes' '' '-work-phone' '(02) 1234 5678' '-ignore-nonexistent-options'
[2017-06-14 08:58:34.619|19576] INFO: HTTP status code is: 500
[2017-06-14 08:58:34.620|19576] INFO: ExecCliGate::InternalServerError[f9fa694d-364f-4780-8a43-7782cb7b3235]: Internal server error: <cli><failure>MySQL query failed: Table 'horde.turba_objects' doesn't exist</failure></cli> [./cmd_exec.cpp:344]

On the target Plesk server, which is a brand new installation, we don't have Horde webmail installed, just Roundcube. I went and manually created horde database and restored the SQL dump fetched from the source onto the target and this appears to have fixed the loop, shortly after the the migration resumed and log entries started to appear again in the info.log file.
Looks like a bug to me, where a command which shouldn't be executed - /usr/local/psa/admin/plib/api-cli/turboaddr.php - returns 500, which causes the loop.
Does this problem still exist in the latest Panel migrator?

 

[UFHH01]

Hi burnley,

first of all, pls. be informed that you should always consider to use the latest release of a product, so you can be sure that former bugs might have been resolved and feature might have been added, to improve the product.

A question like

Does this problem still exist in the latest Panel migrator?

can only be answered from Plesk - Team - Members, who have access to the Plesk internal bug - tracker.


It is always recommended to migrate servers with the very same components installed, to avoid issues/errors/problems, so the statement:

Looks like a bug to me, where a command which shouldn't be executed - /usr/local/psa/admin/plib/api-cli/turboaddr.php - returns 500, which causes the loop.

should be answered with this recommendation and I wouldn't consider such a described issue to be declared as a bug, but rather as a user misconfiguration, because the recommended requirements to avoid issues/errors/problems are not met.

 

[burnley]

Thanks for replying UFHH01. Sometimes(c) running the latest migrator version isn't an option. In our case we need some customizations to speed up the migration of large batches and ensuring mailboxes consistency for, i.e. I use "--delete" switch for rsync when migrating the emails, this avoids the email duplicates. I also skip the "chown -R popuseropuser /var/qmail/mailnames/[...]" because we're using the same uid in both src and dst.
The latest migrator, I think 2.8.7, has implemented some changes which most appear to be related to logging. Creating a local 2.8.7 fork and merging our changes requires time which I don't have right now because, guess what, I'm too busy running migrations
As for the recommended requirements you mentioned, as long as the migrator reports a warning if it doesn't find Horde in the destination I'm happy with this, the clients will be migrated from Horde to Roundcube. But if Plesk glitches on horde database missing and this causes migration delays, this isn't consistent with the warning solution they chose.

 

[Mr.D]

Hello Burnley,

Thank you for your feedback. We have investigated your issue and have conclusion that this delays caused by attempts to restore Horde address books when Horde not installed on the target server. The software issue PMT-3806 has been confirmed.
About loops that you can see in Plesk Migrator debug log these are API requests to Plesk Backup when it restore backup with hosting settings. Such requests are performed over each seconds until hosting settings not will be restored. This is normal behavior.