Boost Plugin v1.6 Released

[TheWitness]

Thanks for the patch. Time is precious.

TheWitness

[John77]

Hi,
Using 0.8.7b and Plugin architecture 2.1, and boost 1.8 (on xen),

I have now the problem, that if I will show the graphs, the first time, no graphs are displayed, or only some of them.
After a refresh (F5), they are all displayed correctly.

I have the same issue with, and without using boost server.

Can anyone give me a hint?

Many Thanks, John

[eterzella.r.t]

I have the same problem !

[jfarese]

me three

[gninja]

Does the on-demand update not remove the updated entries from the output table?

I'm seeing some errors on the regularly scheduled update, and they all seem to be with graphs that I'm actively watching.

[driden91]

I have a question

Actually i have a server with this config :
- one cpu : Intel(R) Xeon(R) CPU 5110 @1.60GHz
- Raid 1 with 2 disk SAS 73G 10K
- 4Gb memory
- debian 2.6.18-5-686-bigmem
- ext3 Filesystem
- cacti 0.8.7b with spine


poller time for this config :

[dononeil]

I just installed v1.8 on my test and production cacti boxes.... on both boxes if I enable boost the charts stop updating. I'm running spine on the production box, and the php poller on the test box.

here is a screen shot of the graphic... you can see the gaps where boost was enabled, then disabled for a few minutes, then re-enabled.

I'm running 8.7b on SLES 10.2, and here are my boost settings (tried tweaking all sorts):

Enable on Demabd RRD Updating - set
Base time: 12:00am
How Often: 6 Hours
Max Records 1000000
Max Recoreds per pass 500,000 (tried 50,000)
Max Insert String - 64000
Update String length - 8192 (tried 2048)
Max RRD Update Time - 1 Hour
Enable Image Caching - Enabled

[Chrysalis]

Hi.

I too had lots of problems with this plugin listed below.

With boost on demand enabled, The graphs stopped plotting data and were been generated with all 0 readings basically, only with timestamps been updated. After a while it got slower and slower loading graph pages with it taking 30 seconds to generate each graph in the browser. Various errors in the cacti log showing the process hitting resource limits and processing 100s of thousands of rrd records. (I only got 8 graphs).

Then turned off the on demand and only had caching enabled for the cached images, this made the graphs plot again and display fast in browser, however every polling interval a php process started owned by cacti and used max cpu until the php execution killed it (set to 5 minutes), in the log errors showing boost had hit resource limits but at least my graphs were been plotted again, why was the boost running with on demand disabled?

I then updated from svn trunk. Then immediatly in the logs I see a very fast boost run of only 58 records and done in a very short time, more like it, and every poll interval no longer generates the red line errors.

Seems the trunk files need to be put in a tarball for a 1.8 version as 1.7 is messy.

[dononeil]

Where is the boost SVN located?

[cd]

svn://cvs.cacti.net/var/svnroot/cacti_plugins

[khufure]

I've just implemented boost on a few of our cacti servers. Even with a development Cacti server, spotting all the problems can be difficult. A few tips:

- Use the SVN version or suffer horrible pain.
- Performance gain is very, very noticeable. Boost is awesome. One of our servers was struggling under the load of ~30,000 RRDs. That same server now comfortably supports 35,000 RRDs with many features still not optimized.
- Be sure the mysql user you are using has DROP TABLE rights. This is not mentioned anywhere in the docs and should be. Without the proper auth rights you will know pain. It needs the auth for rcaston_<random> tables it creates.
- Boost 'multiprocesser' needs to be disabled by default. Boost server failed over and over again with it enabled. Note: I didn't test rrdupdate as I didn't want to introduce more changes than was necessary.
- This plug-in is not for the faint of heart. Expect problems for the first day or two.

Some tips:

A simple shell script to make sure Boost server stays up if it crashes:

/etc/crontab:
# Make sure the Boost poller is running
*/2 * * * * root /usr/local/bin/cacti_rrdsvc

(not to be confused with /etc/init.d/ version)

-bash-3.00$ ls -lat /usr/local/bin/cacti_rrdsvc
-rwxr-xr-x 1 root root 207 Sep 24 03:58 /usr/local/bin/cacti_rrdsvc
#!/bin/sh
# Check to make sure cacti_rrdsvc is running

export RUNCHECK=`netstat -anp | grep 9050 | grep LISTEN | grep -v grep | wc -l`
if [ "$RUNCHECK" -lt "1" ]; then
/etc/init.d/cacti_rrdsvc restart
fi

When you first implement boost it may be worth it to cron a job to
(Modify the interval to your liking)

<Cacti user cron, e.g. /var/spool/cron/cacti or crontab -e -u cacti>
# boost - force update to disk every 30 minutes
15 * * * * /usr/bin/php -q /$pathto/cacti/plugins/boost/poller_boost.php -f -d > /dev/null 2>&1

I kept this cron entry installed as a safety measure.

[dononeil]

The SVN version in trunk is listed as 1.7.... is this the latest official version that works?

[khufure]

[dononeil]

Ok... I've tried using the SVN version of boost and I'm still having problems.

I've verified that my DB user has full permissions.

I imported the correct SQL stuff

I have this in my crontab:
* * * * * php -q /www/cactitest/plugins/boost/poller_boost.php -f -d > /dev/null 2>&1

I copied the other php scripts into my lib folder

I have the settings for boost set like the attached pic.

The second pic is what I get when I graph...

What am I doing wrong?

My web server has write perms on the rrd files, so I don't need to run the 'server'.

[TheWitness]

It could be that your database needs to be tuned. What happens when you run the poller boost at the shell? Have you considered outputing STDOUT and STDERR to a text file from the cron entry? Is the poller_output_boost table loaded?

TheWitness