Issue [PPP-27541] httpdmng --reconfigure-domains exits with errors on migration

[Bitpalast]

Onyx 17.07 #9, CentOS 7.2 64 Bit
Migrator Extension

On every migration Nginx fails. It stops and does not restart automatically during the migration process because of failed httpdmng result. Tested it several times on a fresh install of Onyx and the latest upgrades for OS components. Reproducable. Example:
-----------------------------------------------------------------
[8 Dec 16 17:04:32] [Migration] Plesk restore report problem:
Execution of /usr/local/psa/admin/sbin/httpdmng --reconfigure-domains DOMAIN.TLD -skip-broken failed with return code 1.
Stderr is
[2016-12-08 17:06:21] ERR [util_exec] proc_close() failed ['/usr/local/psa/admin/bin/nginx_control' '--restart'] with exit code [1]
[2016-12-08 17:06:25] ERR [util_exec] proc_close() failed ['/usr/local/psa/admin/bin/nginx_control' '--restart'] with exit code [1]
Can not reload proxy server:
-----------------------------------------------------------------

Nginx can then be restarted manually without problems.
"/usr/local/psa/admin/sbin/httpdmng --reconfigure-domains DOMAIN.TLD -skip-broken" can also be run manually without problems. The migrated domains are migrated completely and fully operational after migration.

 

[IgorG]

Are you sure that nginx is not disabled on the destination server before migration? Check it with

# /usr/local/psa/admin/bin/nginxmng -s

 

[Bitpalast]

Yes, definitely "Enabled".

I have tested it on another target server, too. That other server is not a fresh install but an upgrade from 12.5.30. Same result. Error message as shown above, resulting in Nginx being halted and all websites becoming unavailable until Nginx is restarted manually. Manual execution of "/usr/local/psa/admin/sbin/httpdmng --reconfigure-domains DOMAIN.TLD -skip-broken" and manual restart of Nginx both succeed immediately after migration.

 

[Mr.D]

Hello Peter,

I cannot to reproduce you issue in our test environment. Please provide more details about servers configuration, php settings, subscription php settings and we will try to reproduce this issue again.
Also you can contact our technical support for detail investigation in your environment.

 

[Bitpalast]

There does not seem to be anything special in configuration. I had tested this with different websites/subscription content. To me it feels like a timing issue. It is either that Nginx restarts are "too fast" after a migration completes or "too many too fast". I remember that in one case I saw an error indicating something like that in the journal; maybe the service has not stopped before it restarts; immediately after the migration process finishes httpdmng reconfiguration works like a charm and Nginx can be restarted, too.

I'll try to reproduce it and collect more log data on it, then will report it here. It will take a while, my days are busy.

Source a:
Plesk 12.5.30, latest microupdate, on CentOS 7.2 64 Bit.
Source b:
Plesk 17.07, latest microupdate, on CentOS 7.3 64 Bit.
Both sources: No special extensions active. Some websites with Wordpress from the App repository, some with manual Wordpress installs, some without Wordpress, just plain PHP/HTML. Just ordinary, simple websites, some without PHP scripts.
PHP set to 7.0 on some, 5.6 on some others, FastCGI on some, FPM on some others.
Some source websites with mailboxes, some without.
Some with Let's Encrypt SSL, some without.
More than enough space and resources on the source machine.

Targets:
Two Plesk 17.07, latest microupdate, on CentOS 7.3 64 Bit.
More than enough disk space, CPU power, RAM, inodes, whatever ... Actually a total overkill of machines. CentOS default configuration with "upgrades" to ulimit values of Nginx, Apache, MySQL, system. Nothing really special, no extra special extensions that would not be standard (e.g. Let's Encrypt). Identical subscription settings of the subscriptions that were migrated.

Sources and Targets have ModSecurity and Fail2Ban enabled, but have whitelisted the IPs of all, sources and targets.

 

[Bitpalast]

New test, new bad result on test machine: No errors shown during migration, Nginx "# service nginx status" showing "Active" after migration, BUT: Nginx not serving anything after migration.

No log entries on it, it simply quit serving files and forwarding requests to httpd. I had to restart Nginx manually to get it back working.

 

[Mr.D]

Hello Peter,

Thanks for your feedback, we have already confirmed this case as the software issue "PPP-27541". It will be fixed in future Plesk updates.