the prerequisite for Zabbix

[jenifer1402]

Hello,
they are the prerequisite for Zabbix , CPU,Memory....?
size of memory, cpu...
thanks for any reply

[harmonica]

http://www.zabbix.com/downloads/ZABB...ual%20v1.6.pdf

[lamont]

16-32GB RAM, dual quad-core procs, 6x15k SAS drives, RAID controller with 512MB of cache will be able to monitor 200+ hosts with lots of detail (items) on the hosts, when properly tuned.

[Kerrygeek]

The specs in the manual seem a little low, or maybe I'm just not used to pushing mine quite that hard. I just set mine up in February so I don't have a lot of experience with it yet but it's doing a fine job for us. I have a Xeon 2.5 gig with 4 gigs of RAM running CentOS 5.2 on RAID5 with SAS drives. My load average varies from .35 to about 1.5 but usually stays between .7 and 1.0.

My dashboard says this:

Number of hosts (monitored/not monitored/templates) 129 83 / 1 / 45
Number of items (monitored/disabled/not supported) 3528 1468 / 2050 / 10
Number of triggers (enabled/disabled)[true/unknown/false] 480 479 / 1 [0 / 202 / 277]
Number of users (online) 11 3
Required server performance, new values per second 22.5017

When I look at Admin/queues I don't ever see anything queueing up so I assume I still have plenty of room to grow. I asked about capacity early on and Alexei said not to worry until I had something like 150 values/second so it should still have plenty of head room.

Hope that helps!
Kerry

[lamont]

After thinking about this more, i'd pay most attention to:

1. Hardware RAID controller with battery backed RAM cache (my 512MB RAID caches make me happy, single SATA controllers with no RAID controller make me sad -- i'm not sure if 256MB on the RAID controller is sufficient, but I'd take that over no RAID or software RAID).

2. CPU. The mysql database is actually CPU hungry. I'd want dual quad-core procs next. 4 cores (dual-dual or single-quad) will work, but my current scalability limit seems to be CPU-limited on mysql.

3. Memory. The more memory the larger you can make the innodb buffer pool size and the more data will be cached in memory which makes a difference in responsiveness for looking at historical data. I'd do at least 16GB, but I'm upgrading some of my servers to 32GB now.

4. Faster drives and more of them. If you've got good RAID hardware this actually seems to be second to CPU limits, but clearly 15k SAS drives will be better than 7.2k SATA drives. I'd at least recommend 10k SAS drives. You want more drives for the extra spindles and I/O rather than for space. You aren't going to be too likely to get a zabbix database larger than 100-200GB (i don't think any of mine exceed 100GB, but i prune my history tables fairly aggressively).

So, somewhere between:

256MB RAID controller
4 CPU cores
16GB RAM
2 3.5" 10k SAS drives in RAID10

and:

512MB RAID controller
8 CPU cores
32-64GB RAM
8 2.5" 15k SAS drives in RAID10

[jroberson]

I envy those server specs! In reality it all comes down to what you will be monitoring. My system works decent and I don't have any complaints with how Zabbix is working with what I have given it. Before I tell you my specs, here is my Zabbix:

Number of hosts (monitored/not monitored/templates) 95 62 / 5 / 28
Number of items (monitored/disabled/not supported) 3137 2085 / 991 / 61
Number of triggers (enabled/disabled)[true/unknown/false] 784 775 / 9 [15 / 79 / 681]
Number of users (online) 4 1
Required server performance, new values per second 58.2275

Here is my Zabbix "server" specs:
CPU = Pentium 4 (hyperthreaded) @ 3ghz
RAM = 768MB
HDD = 1x 40GB @ 7200rpm (IDE)



As you can see, I just pulled that "server" out of the pool of desktops (older desktops at that). I admit that the screens and historical graph updates are a little sluggish but all the data is there. My load(15) stays around 1.5-2.0 (while in the frontend it goes to 2.0-3.5)

Disclaimer: This is really my "test" server that I setup to prove to my coworkers that Zabbix is worth implementing. As it turns out, it works pretty well even on that "low" end of hardware. If you are looking for a test case, you don't need to go all out on the "server". I admit, though, that when I'm ever allowed to fully implement it I'm going for some real server hardware, if I can get the $$.

[bbrendon]

Required server performance, new values per second 58.2275

Here is my Zabbix "server" specs:
CPU = Pentium 4 (hyperthreaded) @ 3ghz
RAM = 768MB
HDD = 1x 40GB @ 7200rpm (IDE)

Wow. Very bold "Values Per Second" for such hardware!! Linux is amazing.

[bbrendon]

So, somewhere between:

256MB RAID controller
4 CPU cores
16GB RAM
2 3.5" 10k SAS drives in RAID10

and:

512MB RAID controller
8 CPU cores
32-64GB RAM
8 2.5" 15k SAS drives in RAID10

what the heck do you think they are monitoring? Either of those is enough hardware to monitor thousands of servers!

[alfiere]

I miss something..

My hw is a
2 CPU 4 core Nehalem E5520 @ 2.27GHz
16 GB RAM
2x450 SAS 15 k RAID 1

Centos 5.3 x86_64 well tuned


Number of hosts (monitored/not monitored/templates) 79 61 / 7 / 11
Number of items (monitored/disabled/not supported) 4669 3796 / 541 / 332
Number of triggers (enabled/disabled)[true/unknown/false] 1522 1135 / 387 [2 / 1 / 1132]
Required server performance, new values per second 135.3472 -

[root@~]# cat /proc/loadavg
2.27 2.20 2.34 1/327 14434


with
27552 mysql 15 0 502m 43m 6712 S 17.7 0.3 10203:58 mysqld

mysql never under 15%

(mysql 5.1.34 innodb)

Again, zabbix queue shows:
Items 5 seconds 10 seconds 30 seconds
ZABBIX agent 327 1221 114



Any suggestions?

Regards,
Alf