On Sep 22, 2:12 am, Álvaro J. Iradier <[email protected]> wrote: > Ok, so after cheking newcron.py I can see the main issue here is that > cron tasks are run in a totally independent process, via subprocess > module, so it's impossible to use threading shared data.
true. that is not a good idea because you may have multiple web2py servers behind a load balancer. I would use a database queue as explained here: http://web2py.com/book/default/chapter/04#Background-Processes-and-Task-Queues > > Has rocket a maximum request timeout? If not, I could use cron @reboot > to launch a request to a special URL where a infinite duration task is > started, so I could use that thread to do the sensor polling and share > data with the other threads using cache.ram, or in-memory sqlite > databases. > > I'll give a try to that approach, any suggestions are appreciated. > > Thanks very much. > > On 22 sep, 08:04, Álvaro J. Iradier <[email protected]> wrote: > > > Thanks Massimo, > > > I forgot to mention I'm using the integrated rocket web server (not > > mod_python, cgi or whatever). So I was just curious if there was a way > > of sharing data between the hard "cron" thread and the thread where > > the requests are run, without needing to serialize informate (picke- > > unpickle or anything). I also thought of using a socket, instead of > > named pipes. > > > Basically, I want to avoid database writing because the application is > > run from a SD card, to avoid wearing. > > > However, if there exists not a way for sharing class instances or > > plain variables between cron and request threads, I'm going to take a > > look at using a database from RAM or from a tmpfs. > > > I was thinking of using ":memory:" in-memory sqlite database, but I > > would hit the same problem: I can't share the "db_in_memory" from the > > cron thread to the request thread. > > > Thanks very much. > > > On 21 sep, 22:28, mdipierro <[email protected]> wrote: > > > > If not and you have a single web2py installation you can use a named > > > pipe > > > >http://stackoverflow.com/questions/1430446/create-a-temporary-fifo-na... > > > > has pros and cons. data can be lost or corrupted if the reading/ > > > writing app crashes. > > > I would use the database because of transcaction safety. > > > > Massimo > > > > On Sep 21, 3:21 pm, Álvaro J. Iradier <[email protected]> wrote: > > > > > Hi, > > > > > I've setup a background worker thread using cron @reboot. The > > > > background worker is polling some serial sensors all the time (there's > > > > a loop in class reading the sensor values). On some moment, the values > > > > are recorded and an alarm bit is set. > > > > > Now, what I need is a way to share this information with a web request > > > > without writing it to the database. I've tried so far using class > > > > variables, and cache.ram, but none of them work. I think the request > > > > is being processed in a clean environment, am I right? > > > > > So, what would be a good way of passingsome information to the request > > > > thread from the background worker class? > > > > > Thanks very much. > > > > > Álvaro Iradier. > >
