content of table poller_output available to plugins

gilles · Post by **gilles** » Mon Nov 20, 2006 7:32 am

i want to be able to use the data that poller write into "poller_output" table in a plugin.

the problem is that this table is truncated just before plugins are launched.

is it possible to truncate this table at the start of polling, not at the end ?

the code in question is in cacti/lib folder, poller.php file, i think

fmangeant · Post by **fmangeant** » Mon Nov 20, 2006 7:36 am

gilles wrote:is it possible to truncate this table at the start of polling, not at the end ?

Hi gilles

if this table gets truncated at the start of polling, nothing will get polled ?

Maybe it should be truncated at the end of plugins runtime (I'm using Weathermap, and it's launched at the end of Cacti's polling).

Post by **rony** » Mon Nov 20, 2006 11:29 am

Jimmy is adding a hook that will allow you to copy the poller_output table before truncation.

cigamit · Post by **cigamit** » Mon Nov 20, 2006 9:47 pm

There already is a hook in the current plugin architecture for this.
poller_output
The current SVN of thold takes advantage of it already, the boost plugin also takes advantage of it.

Post by **TheWitness** » Mon Nov 20, 2006 10:11 pm

Boost me baby!

Boost makes developers of things like Manage/Thold/WMap very happy!!!

TheWitness

gilles · Post by **gilles** » Tue Nov 21, 2006 5:55 am

can i have instructions to have svn of thold and boost ?

Post by **TheWitness** » Tue Nov 21, 2006 6:33 am

Gilles,

Boost is still in development. I will send you RC2. Don't have it in SVN though. Jimmy's got THold under SVN though.

Larry

cigamit · Post by **cigamit** » Tue Nov 21, 2006 7:33 pm

Public access to the SVN isn't currently available for a few reasons I shall keep to myself, but it shall be available again in a few weeks. If you would like to know how the hook works, I will gladly explain that though. I also don't mind emailing you a current copy of thold, it just isn't ready for production use (actually, I think it is, I just haven't tested every 15 times to be sure).

Howie · Post by **Howie** » Thu Nov 23, 2006 7:04 pm

TheWitness wrote:Boost me baby! Boost makes developers of things like Manage/Thold/WMap very happy!!!

TheWitness

Boost makes developer of WMap hold his head and try to figure out a way to squeeze it into the recently rewritten-to-be-LESS-cacti-specific WMap, actually, but close

I suspect that other design decisions I've made will actually make WMap+Boost either slightly clunky or slower (for WMap). The Win from Boost will cover for it, hopefully

I still don't fully follow how delaying the rrdtool updates til later actually improves performance though - the writes still have to be done, and things I've been reading lately on rrdtool-users suggest that rrdtool is pretty quick at updates (limited by disk i/o, which it always would be). Obviously the poller returns quicker, but some other process doing the update takes the same CPU and i/o, and if it does it slower than the poller produces them, then it'd never have worked anyway...

I'm not an expert, and don't play one on TV, even. Just curious! It obviously does work, otherwise people wouldn't be so cheerful about it with their giant installations.

cigamit · Post by **cigamit** » Thu Nov 23, 2006 8:15 pm

(limited by disk i/o, which it always would be)

Granted, I'm no expert either, but that is exactly it. Being that it doesn't have to waste so much time with the hard drive seeking to the new rrd file to write 1-2 bits of data. Instead of 10000 seeks every 5 minutes, you can have 10000 seeks every 4 hours (saving 470,000 seeks). It takes almost the same amount of time to write the 1 bit of data, as it does the the 192 bits at once, so the write itself is not the part slowing you down.

Cacti

content of table poller_output available to plugins

content of table poller_output available to plugins

Re: content of table poller_output available to plugins

Who is online