Submit Your CMD.PHP vs. SPINE Metrics Here

[alvaro]

05/12/2006 04:20:15 PM - SYSTEM STATS: Time:134.5436 Method:cactid Processes:4 Threads:2 Hosts:65 HostsPerProcess:17 DataSources:774 RRDsProcessed:490

[JJX]

Debian linux
Intel 2,4 ghz + 512 RAM
Seagate 80Gh hdd

05/17/2006 07:06:35 PM - SYSTEM STATS: Time:3.4218 Method:cmd.php Processes:1 Threads:N/A Hosts:16 HostsPerProcess:16 DataSources:159 RRDsProcessed:147

[kanada]

Hi, some stats

Linux on Intel pIV 2,4Ghz + 512MB ram +160GB HDD

06/20/2006 03:27:31 PM - SYSTEM STATS: Time:144.7878 Method:cactid Processes:6 Threads:1200 Hosts:858 HostsPerProcess:143 DataSources:3401 RRDsProcessed:2206

Linux on 2x Intel pIII 666 + 512MB ram +2x8GB SCSII
&&
Linux on 2x Intel pIII 1,4Ghz + 1024MB ram +4x36GB SCSII

06/20/2006 03:28:22 PM - SYSTEM STATS: Time:201.7262 Method:cactid Processes:4 Threads:8 Hosts:140 HostsPerProcess:35 DataSources:462 RRDsProcessed:358

Seems that pIV win with dual pIII...
I will try to monitor about 10 000 hosts/40 000 DataSources..

Maybe someone have idea what machine will be good for that?

I can buy 2xXEON 3,06Ghz/500Mhz +2GB ram +2x72GB SCSII.. but i dont know that will be good for that.

PS. i can have get repond time <5 min.

Thanks & Greetz,

Alexander.

[gandalf]

See e.g. post at http://forums.cacti.net/viewtopic.php?p=51346#51346 for a rather big installation
Reinhard

[fmangeant]

I upgraded my "old" P4 2.8 / 1 Gb / 2x36 Gb RAID1 to a quad Xeon MP 1.5 / 6 Gb / 2x73 Gb RAID1 + 3x36 Gb RAID5 and here's the result :

Code: Select all

07/05/2006 08:07:11 AM - SYSTEM STATS: Time:129.6936 Method:cactid Processes:6 Threads:17 Hosts:659 HostsPerProcess:110 DataSources:11921 RRDsProcessed:8545

Before the upgrade I had a LOT of gaps in the graphs, and the polling was near 300 seconds.

Next step is to upgrade to latest cactid (I'm still using 0.8.6d because of some crappy Nagios NRPE scripts)...

[wriley]

Dell PE 2850

Dual Intel(R) Xeon(TM) CPU 3.20GHz 2MB Cache
4GB RAM
RAID 1 SCSI 73GB
Gentoo Linux (2.6.16-gentoo-r9)
CACTID 0.8.6f
Cacti 0.8.6h

I am polling every minute, but right now only 3 of the hosts use that interval in the rra and the rest are 5 minute.

Jul 7 09:20:26 cactistl1 Cacti[14557]: SYSTEM: STATS: Time:24.4053 Method:cactid Processes:8 Threads:64 Hosts:298 HostsPerProcess:38 DataSources:37241 RRDsProcessed:12640

[kanada]

Dell PE 2850

Dual Intel(R) Xeon(TM) CPU 3.20GHz 2MB Cache
4GB RAM
RAID 1 SCSI 73GB
Gentoo Linux (2.6.16-gentoo-r9)
CACTID 0.8.6f
Cacti 0.8.6h

I am polling every minute, but right now only 3 of the hosts use that interval in the rra and the rest are 5 minute.

Jul 7 09:20:26 cactistl1 Cacti[14557]: SYSTEM: STATS: Time:24.4053 Method:cactid Processes:8 Threads:64 Hosts:298 HostsPerProcess:38 DataSources:37241 RRDsProcessed:12640

I have machine with something like yours parms..
but on 120 hosts get time request about 200s.

I can monitor about 10000 hosts, so i need best settings to that...

Can you post there you php/mysql/cacti settings?

Thanks,
Alexander

[TheWitness]

Alexander,

What is your mix of snmp, vs. scrips, vs. script server? (poller_output, actions 0,1,2 respectively)

Are you using Cactid or cmd.php?

If you are using cmd.php, do you have php_snmp loaded?

[kanada]

Hi TheWitness,

thank you for answer.. My hardware/system info.
Week ago i buy new server HP:
2xXeon 3.0Ghz
4GB RAM ECC
2x72GB SCSII 15k

System:
Linux 2.4.32
Apache/2.0.54
PHP 5.0.4
mysql Ver 14.7 Distrib 4.1.12
net-snmp 5.2.1.2
php-snmp installed
RRDtool 1.2.11
cacti0.8.6h + cactid0.8.6g

In past my test was great on cactid mode, so now i use it too..
Resoults are not good, on 300hosts load average jump to 20~25 and Time remain: 120.322.. i can only dream how box server i need to request 10 000hosts..
I see high number of cpu/ram eat mysql processess.. every request-time.. must be a better way to request that number of hosts.. in little time, max 300sec.
Im not advance to write that new way, but maybe with good configuration settings, maybe on cmd.php will be better?
Maybe you TheWitness can help me.. of course i can write there all test resoults, as always.

Thanks,
Alexander.

[gandalf]

The answer to the scond question (poller_output table) is missing and cruical for advices. I know a fairly large installation, mainly SNMP polling, with much better results. But scripts (depending on their content) and script server will influence the results heavily.

Personally, I've had some other effect, too. Going public of the cacti site had some effect for us, too, as the number of users increased rapidly (some of them only opening the website for the whole of the day without navigating ...)
Reinhard

[kanada]

Hi again

in this week i modify a bit my cacti instalation.. timings..

07/24/2006 10:05:24 AM - SYSTEM STATS: Time:323.3858 Method:cactid Processes:8 Threads:100 Hosts:4631 HostsPerProcess:579 DataSources:25811 RRDsProcessed:14751

I think that is better but still to high for me..
At this time i send request about traffic interface && script php with 8 parms/oids to every host. Max load average:3.21

Box have something problem with ask snmpget locally.. down screen.. && actual poller settings.
On cacti.log get many errors..
07/24/2006 10:13:52 AM - CACTID: Poller[0] WARNING: SS[8] The PHP Script Server did not respond in time and will therefore be restarted
07/24/2006 10:13:53 AM - CACTID: Poller[0] Host[2104] DS[7108] SS[8] WARNING: Result from SERVER not valid. Partial Result: ...
07/24/2006 10:13:54 AM - CACTID: Poller[0] Host[3276] ERROR: HOST EVENT: Host is DOWN Message: Host did not respond to SNMP
07/24/2006 10:13:57 AM - CACTID: Poller[0] WARNING: SS[6] The PHP Script Server did not respond in time and will therefore be restarted
07/24/2006 10:13:58 AM - CACTID: Poller[0] Host[2118] DS[7164] SS[6] WARNING: Result from SERVER not valid. Partial Result: ...
07/24/2006 10:13:58 AM - CACTID: Poller[0] WARNING: SS[0] The PHP Script Server did not respond in time and will therefore be restarted
07/24/2006 10:13:58 AM - CACTID: Poller[0] Host[2104] DS[7109] SS[0] WARNING: Result from SERVER not valid. Partial Result: ...


Maybe someone will have any sugestion?

Greetz,
Alexander

[ChrisBeeson]

Slackware Linux
2 X 1.6Ghz Processors
4 Gb ram
SYSTEM STATS: Time:292.8269 Method:cmd.php Processes:8 Threads:N/A Hosts:59 HostsPerProcess:8 DataSources:6177 RRDsProcessed:1453

Using cmd.php

I have still got about another 300 devices to add. I assume using cactid will be much faster, I will try that and post the results

[gandalf]

Slackware Linux
2 X 1.6Ghz Processors
4 Gb ram
SYSTEM STATS: Time:292.8269 Method:cmd.php Processes:8 Threads:N/A Hosts:59 HostsPerProcess:8 DataSources:6177 RRDsProcessed:1453

Using cmd.php

I have still got about another 300 devices to add. I assume using cactid will be much faster, I will try that and post the results

Yep. Recommendation for your machine would be to use no more than 4 processes and about 15-20 threads with cactid. We're polling sth like 4 times of your ressources in about 40 sec. time with cactid (most snmp which speeds things up), using a machine with a bit more horsepower
Reinhard

[ruud]

System used :

HP D530 desktop PC
- P4 3Ghz
- 2Gb ram
- Simple IDE HD

Network consisting out of nortel PP8600 and Juniper M320/T640 equipment spread all over the country in 13 regions.

Currently monitoring 389 devices, a total of 11850 graphs..

Polling done by Cactid using 10 concurrent processes, max 10 threads per process, 10 scripts servers and max 60OID's per SNMP Get request.

Poller time currently holding at 32sec average... I'm very impressed.

[gandalf]

Polling done by Cactid using 10 concurrent processes...

Recommended number of concurrent processes is 2* number of processors. So 10 looks a bit insane ...
Reinhard