GoodBye Zabbix

[manowar]

Dear Zabbix,

Oh boy, you cant say I didn't try. I really really did. My whole team really really did.

I picked up on you from a Linux mag interview back in '05. Started using at 1.4 in production in '07. Eagerly awaited escalations in 1.6, and some needed improvements to DM in 1.8. It's not like I'm a newbie.

But things here aren't the same as the simplistic solution back in 2005. And as my infrastructure, and your aims and code, have grown, so have our problems my dear.

Distributed Monitoring. Ahhh DM. This love/hate relationship was always going to end in flames. Promising so much, and never quite delivering. Our first dalliance as a two node parent/child setup was a fun, fruitful time (aside from some minor issues if the child node has an ID > parent ID), but we've never got past that first base. three nodes, four nodes, all still leave two functioning fine, and the rest in a sort of limbo.
The lure of your centralised monitoring screen is fine, but really, have I ever been able to trust all those green squares? Is my far east server quiet because everything is fine, or is it just not sending anything? (Nope, its quiet because Zabbix is asleep on the job. No updates to the master, no emails out. After the first 5 hour outage unreported by zabbix, I'm dissapointed. After another 5 hour outage is reported by my customers not my monitoring, that's when I start to question our affair). So DM, temptress that you may be, you're just a nice idea that never really works. The Lindsay Lohan of monitoring?

And bugs. You are covered in bugs. Escalations that won't stop. Escalations that won't start. Escalations that jump right to the second step (ignoring all my oncall team, you just want to wake me at 2am you saucy tart). Issues with screens, issues with graphs (DM you make me feel dirty). I cant ack alerts from my DM nodes from their parents, I don't see alert data from the nodes. Why do they hide from me?

I promised I wouldn't go on and on.

I don't drop the five years with you lightly. As I said, I've given this my all. I've reinstalled, rebuilt, reconfigured, changed, upgraded until I'm blue in the face. I can compile, RPM, install and configure zabbix in my sleep. And my nightmares.

Bottom line? I can cope with the bugs on their own, they're an annoyance. I'm pretty pissed, but I can kinda live with the fact that the master monitoring screen only shows data from the child nodes, and I have to go to them direct to ack anything. Its another annoyance. But when I cant really, really, be 100% sure that what's shown there is accurate, I start to lose faith. And today, another very significant (and I mean server and all its' services totally dead. Even Servers Up would get this) outage reported to me by my users, not my monitoring, then we reached the bottom of my barrel of faith, perseverance, and tolerance.

Yeah, it looks lovely, it has great graphing and trending, the UI is fantastic but deep down, deep down at its heart, zabbix needs to monitor my kit, and tell me when its broken. Everything else is window dressing. And it just doesnt. Zabbix, FAIL.

I'm done for now. I'm going to rest my head on the pillow of the bitch whore Nagios. I'll pop back now and then to see how you are. Maybe one day you'll grow up and fulfil all those promises you made. But for now, sweet Zabbix, farewell.

[zabbix_zen]

Well, that was....
classy?

Maybe a welcomed improvement, but which would shift away from the current internals,
Zabbix (hard?)adding a built in item+severity high Trigger for each new host creation, alerting by default if no new values are received from it for 20min,
and changing this threshold to 2.5x times(or some similar proportion) the shortest Update Interval of all the items belonging to it.
(would be computed at a new Template addition and recalculated daily?)

Unfortunately the Escalations still behave frailly in some situations..

[nelsonab]

I have to agree, that has to be the most eloquent frustrated goodbye I have ever read.

Anyone who has used Zabbix over the years will agree that it has come a very long way, and continues to do so with each release. What's remarkable is that Alexei gives this software away for free!

With that said, did you consider getting a support agreement? Have you done a price comparison between Zabbix with support and Hyperic or other products? As Alexei has said customers are the biggest drivers of the product, the community is the other main driver.

I'm sorry to hear of your new Nagios pain, I wouldn't wish that program on my worst enemies....

[zabbix_zen]

Zabbix indeed improves a lot each new Release and I hope it will continue to do so and the community keeps helping at that !

However I agree manowar's
remote hosts dead but Zabbix oblivious to it monitoring pain should be avoided in the future for Zabbix newcomers and veterans alike.
An addition of a built in check for this (unfortunately) common scenario would save tons of extra Trigger configuration .nodata() precautions for any Zabbix admin,
keeping newcomers from ever learning this headache even existed.

Nagios... arggggh... wouldn't also be looking forward for that nightmare.

[manowar]

Dont get me wrong, Alexei is doing a fantastic job. Zabbix is really going places. Seeing the evolution from that slightly clunky 1.2 setp all those years back is like watching magic!

We never really considered support. Primarily because it never really worked properly in the first place, so is this an implementation problem? Possibly. We used DM but it just couldnt scale past two nodes with any real reliability. And reliability has to be key in monitoring.

Maybe the answer is that DM just isnt there, and pull it out totally, and run seperate servers in each location. Indeed, given we have to visit them to ack or verify the info from the parent, thats essentially where we are at anyway. But it promises so much, and being able to graph servers from the different environments against each other is really a huge benefit.

I dont want to have to ditch zabbix (and in fact I wont, the graphing and trending is fantastic). And as for the devils daughter Nagios. Say no more. But its a relatively quick deployment and as I said, I *have* to have something reliable now, I can try and iron out the kinks in Zabbix later.

I may have "hammed up" the goodbye letter a tad, Zabbix really is a good project, but for what we've been trying to do with it (and this is nothing that exceptional, the product is sold on the abilities of this) its just failed to deliver, which is sad.

I'll try again, on a split setup (or a DM setup without any monitoring). After I spend the next week editing .cfg files....

[xaviercardoza]

Passion and more Passion

Hi there,

I am new to

- SUSE
- ZABBIX
- Network
- PHP
- MYSQL
- In fact the whole damm lot

But Zabbix is what I need, I have experienced some issues but I love the way I ask a question and I get an answer that helps me on my way and I don't always get the answer I want but the answer I'm given gets me going.

I also believe with all systems/solution they must evolve and with good community it will get better.

So thank for all of the developers for Zabbix and the user community. And for anyone new to this technology stay with it because in life there will always be issues but its what do to fix them or get around them that make it different

[nelsonab]

Hi there,

I am new to

- In fact the whole damm lot

But Zabbix is what I need, I have experienced some issues but I love the way I ask a question and I get an answer that helps me on my way and I don't always get the answer I want but the answer I'm given gets me going.

I realize you have some questions and want some help but it is generally considered bad form and rude to hijack a thread. If you have a question relating to what is going on in the thread, great, please join in! If not, please start a new thread. If you don't know where I would suggest starting with the general help forum here: http://www.zabbix.com/forum/forumdisplay.php?f=6.

Thank you, I hope you find answers to your questions and come up to speed with what you need to know quickly.

[xaviercardoza]

Sorry

Really sorry about that. I did not know. Will not do that again. I was just singing the praises of the application and to give support.

No problem. I understand now, and have learned a valueable lesson.

[nelsonab]

Really sorry about that. I did not know. Will not do that again. I was just singing the praises of the application and to give support.

No problem. I understand now, and have learned a valueable lesson.

It's ok, I think I overreacted a little too. :-) Sorry about that.