Command line Persistent DBI connections

hakkr has asked for the wisdom of the Perl Monks concerning the following question:

Replies are listed 'Best First'.
Re: Command line Persistent DBI connections by perrin (Chancellor) on Dec 13, 2004 at 16:10 UTC
Connecting once per run via connect_cached() doesn't sound so terrible to me, but you could also just have your cron trigger a mod_perl handler that does the work by making an LWP request to your local server.	[reply]
Re: Command line Persistent DBI connections by mpeppler (Vicar) on Dec 13, 2004 at 16:18 UTC
I agree with perrin - connecting once per cron job seems OK to me. The startup time of each cron job will likely dwarf the time it takes to establish the connection. If opening a new connection for each cron job really is a problem then I think you should consider daemonizing some of those cron jobs anyway. Michael	[reply]
Re: Command line Persistent DBI connections by Thilosophy (Curate) on Dec 14, 2004 at 06:06 UTC
I agree with everyone that connecting once at the beginning of every cron run (that would be once a minute at the most, right?) is not an issue. If for some reason you really want a persistant Perl environment for cron jobs, have a look at PPerl on CPAN. This way, you can keep a database connection (and other things you might want to cache). But again, connecting every time seems okay. Since you are not using the connection in between cron runs, keeping it open just wastes resources on the DB server.	[reply]
Re^2: Command line Persistent DBI connections by hakkr (Chaplain) on Dec 14, 2004 at 10:41 UTC
pperl looks very interesting, mod_perl for command the command line. can you use that to test modperl scripts? the reason i ask about the crons is if you have a lot of web servers all running multipe crons against a single db then a lot of connections can start to occur. I suppose the frequency of the crons in question will dictate the need for persistance. if it they are every minute it might be worthwhile cheers	[reply]
Re^3: Command line Persistent DBI connections by Thilosophy (Curate) on Dec 15, 2004 at 05:05 UTC
pperl looks very interesting, mod_perl for command the command line. can you use that to test modperl scripts? Not really. If your modperl script is just "CGI on steroids" (still using the standard CGI interface) you test it by running it as plain CGI. If your modperl script is making use of Apache's internal interfaces, you have to have mod_perl to test it. the reason i ask about the crons is if you have a lot of web servers all running multipe crons against a single db then a lot of connections can start to occur I would say having a lot of crons on a lot of servers is rather a reason to not have persistant connections, because then you have a lot of persistent connections hanging around mostly idle. If you want all your cron jobs (on all servers) to share the same connection, you could move the logic into a mod_perl script that you poll via LWP as suggested by Perrin. This will probably not work very well if the query takes a lot of time. How many servers are we talking here? Go with new connections every time. Chances are your database can handle it just fine. The actual queries are likely to place more of a burden on it than establishing the connections anyway.	[reply]
Re^4: Command line Persistent DBI connections by hakkr (Chaplain) on Dec 17, 2004 at 10:08 UTC


laziness, impatience, and hubris
	PerlMonks