The issue is probably related to opening a subprocess in uWSGI, not the external folder into which you are copying files. Have a look at https://stackoverflow.com/a/17607704. Maybe setting "--close-on-exec" in your uWSGI configuration will work. Alternatively, consider using the web2py scheduler to handle running the task.
Anthony On Wednesday, September 20, 2017 at 7:54:24 AM UTC-4, Lisandro wrote: > > Hi, I'm having a hard time to figure out why web2py throws this error in > this scenario. > > I have one web2py installation, with the same app installed **a lot** of > times. The app is a simple app I've designed to create and mantain your own > website. So the app is one, but it's installed several times, each time > it's installed with a different name, and a specific configuration file > that tells the app wich database to connect, wich domain to serve, etc. > > Up to here, everything works ok. I take the opportunity to say that web2py > really rocks, it's completely scalable! I'm using nginx + uwsgi + web2py. > My whole enviroment receives thousands of requests per day (of course, I > had to consider some tunnings as compile the apps, using lazy_tables, > disabling migrations, etc). But I wanted to say that, because in the past > I've seen some posts saying that web2py wasn't scalable. Nothing more far > from the truth. Not only I use web2py in a high load environment, but also > I developed another web2py app to help me with the administrative tasks of > managing all the other apps. I could successfully develop a web2py app that > allows me to install other apps, change nginx configuration, restart uwsgi, > creating/deleting databases, even upgrading the other apps. I say it again: > web2py really rocks!! > > Back to the issue. Everytime I install a copy of my app in that one web2py > installation, I do it cloning a Mercurial repository I have on > bitbucket.org (by the way, I do all that from my cutom administrative > web2py app). > The problem was that this approach would depend on the availability of > bitbucket.org: if the service is down or degraded, I couldn't perform an > update of the apps. So, I thought I would need to have a local copy on the > repository, and do de pull & update from that local repository, so I > wouldn't depend on bitbucket.org. Every time to time, I would need to do > a pull from the bitbucket repository in order to have the local one up to > date. > > My web2py folder structure looks something like this: > > web2py/applications/administrative > web2py/applications/app1 > web2py/applications/app2 > web2py/applications/app3 > web2py/applications/app4 > web2py/local_repository > > > Notice the administrative application, the installed apps, and the > "local_repository" folder having the cloned repository from bitbucket. From > the administrative app, when I need to add a website, I run a custom code > that clones the local repository as a new app, creates the database, etc, > restarts uwsgi, and reloads nginx configuration. Up to here, it's all good > too. > > Now, the problem appeared when I added a simple function inside the > administrative app, in order to do a pull from bitbucket to the > local_repository. The funcion is very simple, and it uses subprocess to run > an hg pull, as simple as that. And it works ok. The problem is that, after > doing that pull, *all the other apps start throwing errors, and I need to > restart uwsgi in order to avoid those errors*. My first thought was that > restarting uwsgi was needed in the same way is needed when I update an > installed app (that is obvious, if I update an app and compile it, I have > to restart uwsgi in order to make the new code visible). > But, considering that doing a pull is trivial and does not affect the > installed running apps, I wanted to avoid restarting uwsgi in that case. So > my first idea was to move the "local_repository" folder out of the web2py > folder, so my structure looks like this: > > folder/web2py/applications/administrative > folder/web2py/applications/app1 > folder/web2py/applications/app2 > folder/web2py/applications/app3 > folder/web2py/applications/app4 > folder/local_repository > > Notice that now the "local_repository" is completely outside the web2py > folder. > However, and *this is the weird part: after doing a pull inside that > local_repository folder, I still need to restart uwsgi, because if I don't, > all the installed apps start throwing error!* > I'm confused about that. *I don't see how the apps need uwsgi to be > restarted after doing a pull in a folder that is completely outside > web2py's folder*. > > Any thoughts on this? > Thank you very much in advance! > -- Resources: - http://web2py.com - http://web2py.com/book (Documentation) - http://github.com/web2py/web2py (Source code) - https://code.google.com/p/web2py/issues/list (Report Issues) --- You received this message because you are subscribed to the Google Groups "web2py-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
