Speeding up Cacti Poller with Boost

[ukzerosniper]

Hi Guys,

I have just enabled Boost. here is the output from my Boost Status

Code: Select all

Current Boost Status
Boost On Demand Updating:	Idle
Total Data Sources:	1405
Pending Boost Records:	27258
Archived Boost Records:	0
Total Boost Records:	27258
Boost Storage Statistics
Database Engine:	MEMORY
Current Boost Tables Size:	14 MBytes
Avg Bytes/Record:	556 Bytes
Max Record Length:	110 Bytes
Max Allowed Boost Table Size:	14 MBytes
Estimated Maximum Records:	26696 Records
Runtime Statistics
Last Start Time:	2013-03-14 8:58:49
Last Run Duration:	0 seconds (0% of update frequency)
RRD Updates:	0
Peak Poller Memory:	15.3 MBytes
Detailed Runtime Timers:	RRDUpdates:27279 TotalTime:2 get_records:0.7 results_cycle:1.25 rrd_filename_and_template:0.55 rrd_lastupdate:0.04 rrdupdate:0.18 delete:0.06
Max Poller Memory Allowed:	1024 MBytes
Run Time Configuration
Update Frequency:	2 Hours
Next Start Time:	2013-03-14 10:58:49
Maximum Records:	100000 Records
Maximum Allowed Runtime:	20 Minutes
Boost Server Details
Server Config Status:	Enabled
Multiprocess Server:	Single Process
Update Timeout:	2 Seconds
Server/Port:	localhost@9050
Authorized Update Web Servers:	127.0.0.1
RRDtool Binary Used:	/usr/bin/rrdtool
Image Caching
Image Caching Status:	Enabled
Cache Directory:	/var/www/html/plugins/boost/cache/
Cached Files:	39 Files
Cached Files Size:	1 MBytes

I think I have everything setup correctly.

Can someone explain a couple of things to me.

With boost running as I currently have it configured, will my graphs be updated to view as and when I need to look through them, or will they only be available up to the most recent Boost cycle which is set to occur every 2 hours?

If they will only be available after a boost cycle, can I manually update them if I need to see something?

Thanks guys.

[ukzerosniper]

Since activating boost as described in the above post, our graphs have loads of gaps in as shown in the attached screenshot.

Can anyone tell me where I am likely to be going wrong?

Thanks

Code: Select all

http://forums.cacti.net/download/file.php?mode=view&id=27438&sid=e5ecd22ec2327df090cd8a64a04d0440

[ukzerosniper]

Bump

[dtr]

If you enable the "On Demand RRD Updating" option, the graphs will be updated when they are displayed, otherwise I think they will just be updated every 2 hours if that's your interval.

as for the gaps, I used to have them when boost did overrun (takes longer than your update interval to complete). In my case the problem was solved by granting php more memory.

is your output directly after you first installed the boost plugin? Your last run statistics kind of irritate me.

[ukzerosniper]

If you enable the "On Demand RRD Updating" option, the graphs will be updated when they are displayed, otherwise I think they will just be updated every 2 hours if that's your interval.

as for the gaps, I used to have them when boost did overrun (takes longer than your update interval to complete). In my case the problem was solved by granting php more memory.

is your output directly after you first installed the boost plugin? Your last run statistics kind of irritate me.

Hi Dtr,

I had "On Demand RRD Updating" on, I have disabled it now to see if that resolves the gap issue.

My PHP memory limit is set to 756MB at the moment. I had to increase it so that php weathermap could function correctly.

Yeah the output was directly after the first plugin install. Here is the output after having BOOST running for a few days at a 2HR interval.

Thanks.

Code: Select all

Current Boost Status
Boost On Demand Updating:	Disabled
Total Data Sources:	1405
Boost Storage Statistics
Database Engine:	MEMORY
Current Boost Tables Size:	0 Bytes
Avg Bytes/Record:	0 Bytes
Max Allowed Boost Table Size:	14 MBytes
Estimated Maximum Records:	0 Records
Runtime Statistics
Last Start Time:	2013-03-18 13:20:41
Last Run Duration:	2 seconds
RRD Updates:	27296
Peak Poller Memory:	41.92 MBytes
Detailed Runtime Timers:	RRDUpdates:27296 TotalTime:3 get_records:0.69 results_cycle:1.52 rrd_filename_and_template:0.44 rrd_lastupdate:0.1 rrdupdate:0.6 delete:0.03
Max Poller Memory Allowed:	1024 MBytes
Run Time Configuration
Update Frequency:	N/A
Next Start Time:	2013-03-18 15:20:41
Maximum Records:	100000 Records
Maximum Allowed Runtime:	20 Minutes
Boost Server Details
Server Config Status:	Enabled
Multiprocess Server:	Single Process
Update Timeout:	2 Seconds
Server/Port:	localhost@9050
Authorized Update Web Servers:	127.0.0.1
RRDtool Binary Used:	/usr/bin/rrdtool
Image Caching
Image Caching Status:	Enabled
Cache Directory:	/var/www/html/plugins/boost/cache/
Cached Files:	0 Files
Cached Files Size:	0 Bytes

[sg8]

I have boost (version 4.0) installed on one of my cacti (v 0.8.7g) servers

Total Data Sources: 35246
Total Boost Records: 473639
Boost Storage Statistics
Database Engine: MEMORY
Current Boost Table Size: 867 MBytes
Avg Bytes/Record: 545 Bytes
Max Record Length: 113 Bytes
Max Allowed Boost Table Size: 2 GBytes

with "Enable direct population of poller_output_boost table by spine", "Enable On Demand RRD Updating" and Image caching set on. With boost ON Poller process time is lower for avg 20s
Due to company policy i need to migrate my cacti to another server, so i tried to install new cacti, new spine. Everything worked well until i enabled the last boost version. Now i get Poller error about overrunning processes. I've noticed that new version of boost doesn't have input field for interlocking file (in the settings). Is it mistake or its the way it shoud be???

[ukzerosniper]

OK Guys,

Since I disabled the on demand updating we don't get gaps in the graphs any more.

However, our cacti poller runtime is showing that we have not reduced the pollers run time at all.

See graph below.



The Boost Status is as follows.

Code: Select all

Current Boost Status
Boost On Demand Updating:	Disabled
Total Data Sources:	1405
Boost Storage Statistics
Database Engine:	MEMORY
Current Boost Tables Size:	0 Bytes
Avg Bytes/Record:	0 Bytes
Max Allowed Boost Table Size:	14 MBytes
Estimated Maximum Records:	0 Records
Runtime Statistics
Last Start Time:	2013-03-18 13:20:41
Last Run Duration:	2 seconds
RRD Updates:	27296
Peak Poller Memory:	41.92 MBytes
Detailed Runtime Timers:	RRDUpdates:27296 TotalTime:3 get_records:0.69 results_cycle:1.52 rrd_filename_and_template:0.44 rrd_lastupdate:0.1 rrdupdate:0.6 delete:0.03
Max Poller Memory Allowed:	1024 MBytes
Run Time Configuration
Update Frequency:	N/A
Next Start Time:	2013-03-18 15:20:41
Maximum Records:	100000 Records
Maximum Allowed Runtime:	20 Minutes
Boost Server Details
Server Config Status:	Enabled
Multiprocess Server:	Single Process
Update Timeout:	2 Seconds
Server/Port:	localhost@9050
Authorized Update Web Servers:	127.0.0.1
RRDtool Binary Used:	/usr/bin/rrdtool
Image Caching
Image Caching Status:	Enabled
Cache Directory:	/var/www/html/plugins/boost/cache/
Cached Files:	75 Files
Cached Files Size:	2 MBytes

Can anyone give me any ideas why the pollers runtime is not reduced?

Thanks all.

[sg8]

I had same issue because i forget to specify lock file for boost (talking about boost version 4.0)..

[ukzerosniper]

I had same issue because i forget to specify lock file for boost (talking about boost version 4.0)..

Did you have the problem that the poller performance was not increasing with boost?

[sg8]

It was always higher that 55s in 1 minute polling time. As I sad i had no file lock specified so there were 2 processes trying to work with the same files so i guess collision was the problem.
After i set file lock i've played a little with number of threads and script servers (i have about 36k of data sources - 16 threads and 5 script servers were enough for my needs) and cacti performances were amazing.

[ukzerosniper]

How do I configure "file lock"?

[sg8]

I'm using boost version 4.0 so when you go to Settings->Boost and there you have

Process Interlocking
Path to lockfile for On Demand RRD Updating
This lock file will assist the boost poller from detecting overrun conditions.

and i set this to /tmp/cacti-boost.lockf

Now, as i mentioned before, in current boost version i don't have this field. I just have Process Interlocking but no field to enter path to lock file..so maybe we should contact plugin author for more information ?

[phalek]

Version 5 uses a MySQL function called "GET_LOCK" to implement this.

http://dev.mysql.com/doc/refman/5.0/en/ ... n_get-lock

Version 4 used a file-system based lockfile.

Hence the header for the process lock in Version 5 is a "left-over" from Version 4. The lockfile as defined in V4 is not needed any more

[sg8]

Muchas graaaaacias

[ukzerosniper]

I found that increasing "Maximum Threads per Process" made a bit improvement. What is the Maximum Threads per Process that you would normally use?