From Zabbix forum - can be interesting from Victor Kirhenshtein on 2004-05-11 (NetXMS-dev)

From: Victor Kirhenshtein <victor_at_DOMAIN_REMOVED>
Date: Tue, 11 May 2004 16:27:20 +0300

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2563451
By: nobody

Hi Alexei,

On The SLA subject again, I think that adding some modifications to the current
module, would make it really usefull :
- business hours and non working days when define triggers
- check triggers dependencies in It services

Having a reporting module to generate some PDF documents will be very appreciated.
All data collected by Zabbix could be "easilly" retrieved and processed using
an external module.

Perhaps some guys can help to develop it ?
Can it be a idea to add an "Reporting Module for Zabbix" to your Zabbix Project
?

Thanks
BD

______________________________________________________________________
You are receiving this email because you elected to monitor this forum.
To stop monitoring this forum, login to SourceForge.net and visit:
https://sourceforge.net/forum/unmonitor.php?forum_id=74298

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2562344
By: hugetoad

Yes, I live in Riga. I was in Vilnius, Palanga, Klaipeda, and Neringa many times.
I like Lithuania very much! :]

I agree that Zabbix should not manage financial aspects of SLA.

The SLA topic was very interesting. I'll put a note into my TODO list to make
sure that it won't be forgotten. The complex SLA functionality has to be implemented
at some point. Obviously it depends on priorities and availability of resources.

Best regards,
Alexei

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2562309
By: nobody

Hi again, Alexei :-)
Yes, I am from Lithuania, small country near Latvia :-)

About expectations - of course it would be excellent if it could do financial
calculations, but I think it is not necessary. Every tool should do the best
it can do except those things it is not intended for :-) I think it good UNIX
philosophy :-)

Seriously, financial calculations should be provided at financial department,
so system administrators nor helpdesk should not do this job.

For example, some contract can be changed and financial things can start to
differ, regardless that monitoring system settings or SLA can still be the same.

So it would be sufficient to have only counting of downtime hours by class,
e.g. "hours overdued in separate cases at working hours", hours overdued in
separate cases at non working hours", "hours overdued per month", etc..

Best Regards - Ricardas.
(Do you Live in Riga? I had been there some time ago, it has beautifull, neat
old town)

P.S.:
To creator of this thread: ("Why Zabbix"):
Really it is still difficult to me to say something about real differences,
but I see at least two important aspects of Zabbix:

1. I think Zabbix would be faster than Nagios (Huh, I still haven't tried it,
so it is speculative opinion, but I will try). When you do some check with Nagios,
it forks new perl process (this operation is processor-greedy). And when you
need to check several hundred things per minute it could generate very heavy
load on monitoring machine. I am not sure, but it seems that Zabbix has bit
different checking implementation, so it should be more efficient.

2. Normally, Nagios is configured by editing several config files and configuration
syntax is not very simple. In some cases it is feature but when you want to
do something easilly, without deep learning, it could be a problem. For example
you can not give permission to configure hosts monitored by Nagios for ordinary
supporter.

(Really, if you want to add one independent thing to Nagios monitoring you may
need to write ~100 lines of text in several configuration files (contacts, checks,
hosts, services, etc.). Other side of this configuration language: if you wrote
some template or general config, you can apply it to new host by writing only
two or three additional strings).

3. Nagios is oriented mainly to alerting somebody in case of something goes
wrong. As alerter it is really good. But when you need to monitor ping times
or processor load, you will get only alerts. So you will need to use poorly
supported Nagios extensions or some additional tools like MRTG.

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2562188
By: hugetoad

Hi Ricardas,

Thanks for detailed explanation. I appreciate it! Actually it is not as difficult
as it sounds. Zabbix lacks two bits of functionality to make it real:

- calendars (24x365, Working Time, Non-working Time, Holidays, Downtimes, whatever
- can be very flexible)
- additional report for complex SLA calculation

I'm not sure what could be best design for this, however I do not see any reasons
why it couldn't be implemented.

By the way, do you expect Zabbix to perform financial calculations as well?

Are you from Lithuania?

Best regards,
Alexei

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2562009
By: nobody

Hi again!
It is quite glad to see such a fast response from you, Alexei, thank you.

I would like to explain what are those "complex availability calculations" :-)
Of course it can be difficult to explain :-)

In commertial world there are some kind of contracts where fines are provided
for unavailability of service. In those contracts (SLA) usually are defined
two (sometimes three or even four) regular periods of service time:
- Working hours (for example, from 09:00 to 17:00 except sundays)
- Non service hours (in this case from 17:00 to 09:00 and 00:00 to 24:00 on
sundays)

Really, "Non service hours" does not mean that there can be no service at that
time. It only means that fines and service restoration times can be different
than at "Working hours".

Alse there can be two lenghts of reaction times:
- simple reaction time (time to inform customer or specialist about the
problem)
- service restoration time (when service starts to work again).

Usually there only one important time: service restoration.

So, when we have contract with customer (called SLA - Service Level Agreement)
there are defined important things, for example:

1. In period of working hours (between 09:00 and 17:00) there will be no service
interruption for time longer than 2 hours. In case of any longer interruption
fine should be paid.
2. Maximum total time of service interruptions at working hours can be 8 hours
per month. In case of longer non service time fine should be paid.
...
So, if we have service interruption for 3 hours, ve need to pay fine. But we
need to pay fine in case of there were only service interruptions for less than
half of hour, but total interrupition time was 9 hours at month.
But if we have one interruption for 3 hours and one for 2 hours we have total
interruption for 5 hours, but again, one of those service interruptions is overdued,
so it should be fined.
...

And it is not all: we can have different schema of fines for non working hours,
and again, there can be defined some kind of "customer's fault", when fine should
not be paid and planned downtime when service availability should be assumed
as "UP" too.

So, it is not easy thing :-)

Anyway, it seems that you have created really useful and interesting tool, IT
services thing seems to be especially interesting (as far as I know, Nagios
does not have this kind of feature implemented) so I will try to set up it at
our company and maybe we will contribute to it somehow :-)

Best Regards - Ricardas Savukynas.

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2561821
By: hugetoad

Currently, support of hundreds of hosts adds quite significant overhead to Zabbix
administrators provided that host information is not static. It is because of
lack of a mechanism which would allow to propagate changes made for a template
to related hosts. This functionality will be available in the next stable release.
Hopefully, it will make possible maintenance of hundreds of servers with several
mouse clicks.

Zabbix does not support "working time" and "downtime" yet. Currently it considers
24x7x365 as working time. There are no end-user oriented detailed SLA report,
however all information for the report is available in Zabbix database. I'm
not sure what "complex availability calculations" is. Zabbix is able to calculate
SLA for everything: whole IT department, databases, specific applications, a
single host, a network interface, everything. See IT Services for more details.

Yes, Zabbix does support agentless monitoring. That's is what simple checks
are about.

My goal is to make Zabbix 1.1 (next stable release) to support monitoring of
medium-sized networks, i.e. networks having up-to 1000 hosts for monitoring.
Zabbix frontend will be optimised and in some cases simplified as well.

Let me know if you need more detailed information...

Best regards,
Alexei

attached mail follows:

Read and respond to this message at:
https://sourceforge.net/forum/message.php?msg_id=2561739
By: nobody

Hi, from documentation it seems that ZABBIX would be quite useful for small
networks, especially when you need to easy configured monitoring system. But
how it useful for big ones?

For example, our company is services provider for hundreds of other companies.
How is it easy to support such system?
(In nagios you can use templates to simply configure class of hosts or services,
and later you can add few strings for new company. In case of something changed
for all the class of companies you can simply change template.)

Also, other question: in most cases you need not only for alert when some host
or service is down but for more detailed service overview later.

For example: you can have some agreement with some company that there will be
maximum downtime of 8 hours per month at working hours and 16 hours at non working
hours at all. And maximum lenght of one downtime will be at most 2 hours at
working hours and at most 8 hours at non working hours.
In case of such SLA broken you will need to pay back. And of course is not easy
to count those hours :) As far as I know Nagios does not support this kind of
calculations, but it is supported by some of commercial monitoring tools, such
as Nimbus. How it is with ZABBIX? Does it support complex availability
calculations?

Also, by the way: does it offer possibility to do monitoring as service? For
example: we can set up nagios to monitor some services at some host for external
company. In this case that host and service information should be available
only for two groups of users: monitoring helpdesk team and company who is monitored
but not viewed by others. Can we do that with ZABBIX?

This archive was generated by hypermail 2.2.0 : Mon Aug 01 2005 - 00:43:15 EEST