[Koha] Plack unable to restart once every week

[Andreas Roussos]

Dear List,

We are running Koha 16.05.02 on Debian 8, installed from packages.

Recently we enabled Plack, and have been observing the following behaviour:
once every week on a Sunday, the Plack process will fail to restart.
This results in a "Service Unavailable" page when trying to load the Staff
Interface.
The relevant entries from the log files are marked with an asterisk below.

/var/log/koha/imp$ zegrep "closing|starting" plack-error.log* | sort -t ':'
-k 2
plack-error.log.5.gz:2017/03/27-06:29:17 Server closing!
plack-error.log.4.gz:2017/03/27-06:29:19 Starman::Server (type
Net::Server::PreFork) starting! pid(26134)
plack-error.log.4.gz:2017/03/28-06:29:17 Server closing!
plack-error.log.4.gz:2017/03/28-06:29:20 Starman::Server (type
Net::Server::PreFork) starting! pid(3699)
plack-error.log.4.gz:2017/03/29-06:29:15 Server closing!
plack-error.log.4.gz:2017/03/29-06:29:17 Starman::Server (type
Net::Server::PreFork) starting! pid(11642)
* plack-error.log.4.gz:2017/04/02-06:29:23 Server closing!
* plack-error.log.3.gz:2017/04/02-11:28:27 Starman::Server (type
Net::Server::PreFork) starting! pid(972)
plack-error.log.3.gz:2017/04/03-06:29:23 Server closing!
plack-error.log.3.gz:2017/04/03-06:29:27 Starman::Server (type
Net::Server::PreFork) starting! pid(5798)
* plack-error.log.3.gz:2017/04/09-06:29:25 Server closing!
* plack-error.log.2.gz:2017/04/09-13:58:48 Starman::Server (type
Net::Server::PreFork) starting! pid(964)
plack-error.log.2.gz:2017/04/11-06:29:30 Server closing!
plack-error.log.2.gz:2017/04/11-06:29:33 Starman::Server (type
Net::Server::PreFork) starting! pid(9831)
* plack-error.log.2.gz:2017/04/16-07:36:00 Server closing!
* plack-error.log.2.gz:2017/04/16-09:02:50 Starman::Server (type
Net::Server::PreFork) starting! pid(984)
plack-error.log.2.gz:2017/04/17-06:29:25 Server closing!
plack-error.log.1:2017/04/17-06:29:27 Starman::Server (type
Net::Server::PreFork) starting! pid(5563)
* plack-error.log.1:2017/04/23-06:29:32 Server closing!
* plack-error.log:2017/04/23-10:36:14 Starman::Server (type
Net::Server::PreFork) starting! pid(7389)

As you can see, on the 2nd/9th/16th/23rd of April (i.e. every 7
days) the Plack process did not restart automatically and we had
to start it manually with `koha-plack --start imp`.

FWIW, we do not have anacron installed and haven't made any
changes to the Koha logrotate script.

Does anyone have any leads as to what may be causing this behaviour?

Thanks in advance!

Kind regards,
Andreas Roussos
Systems Administrator
Holy Monastery of the Paraclete