Submit Your CMD.PHP vs. CACTID Metrics Here

andrew2 · Posted: Fri Apr 15, 2005 10:04 am Post subject:

Just made the switch to cactid. The old php poller was taking ~80 seconds to complete. With cactid:

SYSTEM STATS: Time: 11.0186 s, Method: cactid, Processes: 5, Threads: 5, Hosts: 62, Hosts/Process: 13

maxpower · Joined: 12 Jul 2004 Posts: 3

For cactid my results are typically as follows:
04/20/2005 12:05:20 PM - SYSTEM STATS: Time: 18.9282 s, Method: cactid, Processes: 3, Threads: 4, Hosts: 126, Hosts/Process: 42

the php poller, on the other hand, was as follows when I tinkered with it:
04/20/2005 12:08:27 PM - SYSTEM STATS: Time: 41.3850 s, Method: cmd.php, Processes: 3, Threads: N/A, Hosts: 126, Hosts/Process: 42

So far we have...
Devices: 125
Graphs: 4012
Data Sources: 4230

I still have a lot to add and will update the stats as I get more put in.

ulasyuksel · Joined: 08 Apr 2005 Posts: 5

04/26/2005 04:45:40 PM - SYSTEM STATS: Time: 39.7222 s, Method: cactid, Processes: 5, Threads: 4, Hosts: 88, Hosts/Process: 18
04/26/2005 04:50:43 PM - SYSTEM STATS: Time: 42.9924 s, Method: cactid, Processes: 5, Threads: 4, Hosts: 88, Hosts/Process: 18
04/26/2005 04:55:41 PM - SYSTEM STATS: Time: 41.0732 s, Method: cactid, Processes: 5, Threads: 4, Hosts: 88, Hosts/Process: 18

Darwin · Joined: 27 May 2002 Posts: 27 Location: Montreal, QC

Wow.. You guys weren't kidding when you said polling performance was a big improvement with cactid.

This is on a Dual Xeon 1.8GHz.

Using cmd.php:

(It was often reaching the 300 second threshold, perhaps 1 out of 4 times)

04/27/2005 01:34:01 PM - SYSTEM STATS: Time: 241.3832 s, Method: cmd.php, Processes: 3, Threads: N/A, Hosts: 93, Hosts/Process: 31
04/27/2005 01:59:11 PM - SYSTEM STATS: Time: 251.8699 s, Method: cmd.php, Processes: 3, Threads: N/A, Hosts: 93, Hosts/Process: 31
04/27/2005 02:04:54 PM - SYSTEM STATS: Time: 294.1170 s, Method: cmd.php, Processes: 3, Threads: N/A, Hosts: 93, Hosts/Process: 31

Using cactid:

(Seems putting more processes and also more threads is what helps the most)

04/27/2005 02:27:07 PM - SYSTEM STATS: Time: 126.1569 s, Method: cactid, Processes: 3, Threads: 3, Hosts: 93, Hosts/Process: 31
04/27/2005 02:34:27 PM - SYSTEM STATS: Time: 266.3451 s, Method: cactid, Processes: 1, Threads: 3, Hosts: 93, Hosts/Process: 93
04/27/2005 02:39:46 PM - SYSTEM STATS: Time: 285.7999 s, Method: cactid, Processes: 1, Threads: 3, Hosts: 93, Hosts/Process: 93
04/27/2005 02:44:47 PM - SYSTEM STATS: Time: 286.4331 s, Method: cactid, Processes: 1, Threads: 3, Hosts: 94, Hosts/Process: 94
04/27/2005 02:47:26 PM - SYSTEM STATS: Time: 145.4549 s, Method: cactid, Processes: 3, Threads: 3, Hosts: 94, Hosts/Process: 32
04/27/2005 02:52:30 PM - SYSTEM STATS: Time: 150.6067 s, Method: cactid, Processes: 3, Threads: 1, Hosts: 94, Hosts/Process: 32
04/27/2005 03:07:20 PM - SYSTEM STATS: Time: 138.9387 s, Method: cactid, Processes: 3, Threads: 3, Hosts: 94, Hosts/Process: 32

TheWitness · Posted: Wed Apr 27, 2005 5:48 pm Post subject:

Darwin,

With a dual processor box, I would use 4 processes and 5 to 8 threads each.

TheWitness

mpdsville1 · Posted: Mon May 09, 2005 12:22 pm Post subject: Cactid -vs- Cmd.php

Dual P-III-700s , 2gb ram, Fedora Core3 ,
Cacti 0.8.6c and Cactid 0.8.6d

bash-3.00# ls /usr/local/cacti/rra|wc
1441 1441 49752
All routers and switches.

Significant Improvement is an understatement.

05/04/2005 02:52:50 PM - SYSTEM STATS: Time: 168.7650 s, Method: cmd.php, Processes: 5, Threads: N/A, Hosts: 96, Hosts/Process: 20
05/04/2005 02:57:36 PM - SYSTEM STATS: Time: 154.9236 s, Method: cmd.php, Processes: 5, Threads: N/A, Hosts: 96, Hosts/Process: 20
05/04/2005 03:02:37 PM - SYSTEM STATS: Time: 155.1510 s, Method: cmd.php, Processes: 5, Threads: N/A, Hosts: 96, Hosts/Process: 20

05/04/2005 03:05:14 PM - SYSTEM STATS: Time: 11.4291 s, Method: cactid, Processes: 5, Threads: 5, Hosts: 96, Hosts/Process: 20
05/04/2005 03:10:17 PM - SYSTEM STATS: Time: 15.3489 s, Method: cactid, Processes: 5, Threads: 5, Hosts: 96, Hosts/Process: 20
05/04/2005 03:15:16 PM - SYSTEM STATS: Time: 14.8460 s, Method: cactid, Processes: 5, Threads: 5, Hosts: 96, Hosts/Process: 20

Darwin · Joined: 27 May 2002 Posts: 27 Location: Montreal, QC

Witness,

For some strange reason, if I raise the number of processes up to 4, I seem to run into issues where all the cactid processes seem just "stuck there" doing nothing. The maximum runtime then gets reached, I get output in the log file that 296 secs have been reached, but all the processes stay there, forcing me to have to manually kill them all, otherwise they start accumulating and queing up when the next poller.php runs.

The maximum combination I seem to be able to reach is 3 processes and 4 threads. The same problem happens if I stay at 3 processes and go up to 5 threads.

Another thing, let's say I specify 3 threads and 5 processes, how many simultaneous mysql connections are created? I currently have a limit of 25 processes for this mysql user and am wondering if this might have a cause of this issue...

TheWitness · Posted: Tue May 10, 2005 7:49 pm Post subject:

Each thread will take a connection. Therefore, if you have many users, you might be hitting the wall.

TheWitness

williem · Cacti User Joined: 08 Feb 2005 Posts: 59

Hey,

I have not been able to run cmd.php for a long time as it would go well over the 296 sec range. Here is the latest info from my cactid log.

05/11/2005 09:36:40 AM - SYSTEM STATS: Time: 99.7499 s, Method: cactid, Processes: 4, Threads: 20, Hosts: 2020, Hosts/Process: 505

I am running cacti on a 3ghz dual processor dl580 with 2 gig ram.

Regards,

Willlie

tapufd · Cacti User Joined: 19 Aug 2004 Posts: 55 Location: Belgium

I have for the moment 157 hosts configured, with a total of +- 900 data sources. To poll all data, cacti needs about 150sec. Is that a normal number? Because I find numbers of :
* 5000 data sources in 40 sec.
* About 17 to 18 sec. for cactid, 2 Processes, 8 Threads, 133 Hosts, 1390 rrds
* 3/15/2005 10:11:31 PM - SYSTEM STATS: Time: 91.0645 s, Method: cactid, Processes: 2, Threads: 25, Hosts: 346,Hosts/Process: 173

Seems I'm a lot behind.

I have noticed that changing the number of threads per process doesn't really influence performance. Only changing the number of processes speeds up the process. Is this normal in case of cactid?

/tapufd

using : HP ProLiant DL380 G3; 2x 3Ghz cpu; 2gig of ram; Windows 2K3; cacti 0.8.6d + all patches; cactid 0.8.6e.RC7

elnino · Cacti User Joined: 08 Mar 2005 Posts: 205

Pentium-III 733MHz, 384mb RAM running Linux
Cacti 0.8.6d with Cactid 0.8.6d

Currently:
40 devices - Mixture of Novell and Microsoft servers, switches and routers
303 data sources

Using the PHP poller:
05/28/2005 08:36:13 PM - SYSTEM STATS: Time: 71.4499 s, Method: cmd.php, Processes: 2, Threads: N/A, Hosts: 39, Hosts/Process: 20
05/28/2005 08:31:14 PM - SYSTEM STATS: Time: 72.3977 s, Method: cmd.php, Processes: 2, Threads: N/A, Hosts: 39, Hosts/Process: 20
05/28/2005 08:26:13 PM - SYSTEM STATS: Time: 72.3474 s, Method: cmd.php, Processes: 2, Threads: N/A, Hosts: 39, Hosts/Process: 20
05/28/2005 08:21:13 PM - SYSTEM STATS: Time: 71.3342 s, Method: cmd.php, Processes: 2, Threads: N/A, Hosts: 39, Hosts/Process: 20
05/28/2005 08:16:12 PM - SYSTEM STATS: Time: 71.4442 s, Method: cmd.php, Processes: 2, Threads: N/A, Hosts: 39, Hosts/Process: 20
05/28/2005 08:11:13 PM - SYSTEM STATS: Time: 71.3216 s, Method: cmd.php, Processes: 2, Threads: N/A, Hosts: 39, Hosts/Process: 20
Sometimes it would take 85-90 seconds

Using Cactid:
05/28/2005 09:00:15 PM - SYSTEM STATS: Time: 13.5017 s, Method: cactid, Processes: 2, Threads: 2, Hosts: 39, Hosts/Process: 20
05/28/2005 08:55:14 PM - SYSTEM STATS: Time: 13.3899 s, Method: cactid, Processes: 2, Threads: 2, Hosts: 39, Hosts/Process: 20
05/28/2005 08:50:15 PM - SYSTEM STATS: Time: 13.4738 s, Method: cactid, Processes: 2, Threads: 2, Hosts: 39, Hosts/Process: 20
05/28/2005 08:45:15 PM - SYSTEM STATS: Time: 13.1896 s, Method: cactid, Processes: 2, Threads: 2, Hosts: 39, Hosts/Process: 20
05/28/2005 08:40:14 PM - SYSTEM STATS: Time: 13.3621 s, Method: cactid, Processes: 2, Threads: 2, Hosts: 39, Hosts/Process: 20

havoc1942 · Joined: 26 Oct 2004 Posts: 11

Well... cmd won't complete but here are my cactid stats.

06/08/2005 04:32:18 PM - SYSTEM STATS: Time: 137.4013 s, Method: cactid, Processes: 4, Threads: 50, Hosts: 319, Hosts/Process: 80

Dual P3 1G XEON w/ 3G memory
37,810 rrd's

MySQL running on a seperate server
Uptime 37 minutes
Threads: 13 Questions: 871236 Slow queries: 0 Opens: 682 Flush tables: 1 Open tables: 145 Queries per second avg: 424.993

I have hit the peak of performance with this install and I need more hosts monitored....I don't have any better boxes, with the exception of moving to a quad PIII 733XEON...suggestions?

TheWitness · Posted: Thu Jun 09, 2005 5:25 am Post subject:

havoc1942,

I have limited threads to 30 in the latest version. How is performance affected when you reduce threads to 30? Also, please post the number of snmp, vs. scripts, vs. script_server DS's of those 37k (Action 0, Action 1, Action 2).

Thanks,

TheWitness

JimmyC · Joined: 17 Feb 2005 Posts: 14 Location: Belgium

Dear,

I did some testing with the # processes and # of threads; here are the results.

06/27/2005 10:12:38 AM - SYSTEM STATS: Time: 38.4082 s, Method: cactid, Processes: 8, Threads: 8, Hosts: 103, Hosts/Process: 13, Data Sources 3941, RRDs Processed 2093
06/27/2005 10:07:38 AM - SYSTEM STATS: Time: 38.5363 s, Method: cactid, Processes: 8, Threads: 8, Hosts: 103, Hosts/Process: 13, Data Sources 3941, RRDs Processed 2093

06/27/2005 09:57:49 AM - SYSTEM STATS: Time: 49.7013 s, Method: cactid, Processes: 4, Threads: 8, Hosts: 103, Hosts/Process: 26, Data Sources 3941, RRDs Processed 2093

06/27/2005 09:53:14 AM - SYSTEM STATS: Time: 74.5356 s, Method: cactid, Processes: 1, Threads: 12, Hosts: 103, Hosts/Process: 103, Data Sources 3941, RRDs Processed 2093

06/27/2005 09:48:17 AM - SYSTEM STATS: Time: 76.9617 s, Method: cactid, Processes: 1, Threads: 8, Hosts: 103, Hosts/Process: 103, Data Sources 3941, RRDs Processed 2093
06/27/2005 09:43:17 AM - SYSTEM STATS: Time: 77.1579 s, Method: cactid, Processes: 1, Threads: 8, Hosts: 103, Hosts/Process: 103, Data Sources 3941, RRDs Processed 2093

Machine is Dell PE 2650 ; Dual Xeon 3.0 Ghz ( HT ) ; 4 gb ram.
99% of the datasources are SNMP gets.
Only a couple of scripts ( Perl ).

shanali · Joined: 20 Feb 2003 Posts: 20

07/30/2005 01:42:04 PM - SYSTEM STATS: Time: 123.3016 s, Method: cactid, Processes: 1, Threads: 3, Hosts: 34, Hosts/Process: 34, Data Sources 3149, RRDs Processed 1693
07/30/2005 01:47:22 PM - SYSTEM STATS: Time: 140.8046 s, Method: cactid, Processes: 1, Threads: 3, Hosts: 34, Hosts/Process: 34, Data Sources 3149, RRDs Processed 1693
07/30/2005 01:52:18 PM - SYSTEM STATS: Time: 137.4742 s, Method: cactid, Processes: 1, Threads: 3, Hosts: 34, Hosts/Process: 34, Data Sources 3149, RRDs Processed 1693

on a dell with load averages: 0.19, 0.13, 0.09

FreeBSD 5.4-STABLE #4: Fri Jul 22 10:32:17 SGT 2005
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: Intel(R) Xeon(TM) CPU 2.80GHz (2791.01-MHz 686-class CPU)
Origin = "GenuineIntel" Id = 0xf29 Stepping = 9
Hyperthreading: 2 logical CPUs
real memory = 536674304 (511 MB)
avail memory = 515514368 (491 MB)

running

cacti-0.8.6f Web-driven graphing interface for RRDTool
cactid-0.8.6.e A multithreaded poller for Cacti written in C

i have tried a few combinations of cactid (concurrency/threads) but it just gets worse

...