On Fri, Aug 14, 2009 at 3:32 PM, Derek J. Balling<[email protected]> wrote: > > On Aug 14, 2009, at 9:55 AM, Rob Cherry wrote: >> Anyone hate Zabbix with a flaming passion in favour of anything else? > > We hate Zabbix with a flaming passion. Here's cool features you'll love:
I hate it, but not with a passion, and not any more than I hate every other monitoring app I've used. I've had problems with the Zabbix server freezing up, so that it stops collecting data, stops processing alerts, and just sits there doing it's tree impression. This used to happen several times a day. The number of server and client processes that in requires is also ridiculous, in my opinion. However, it's ability to read any data that you can collect in a script is very useful; it also has the most feature full interface of anything I've used. (Even if it is a royal pain to figure out how those features work.) > -- If you create some aggregate graphs (e.g., "the load averages for > all the web servers in my farm")... and you retire web10, that > aggregate graph? Yeah, it's gone. Zabbix's attitude appears to be > "nuke the site from orbit, it's the only way to be sure", and will > destroy and graphs/screens which include that host in it. (At least > that seems to be the experience of our folks who work more closely > with maintaining our much-hated Zabbix install). I Haven't seen this, but from what I hear this was recently fixed. Zabbix development is active, and every update is worth it. > -- It does NOT scale well at all... we've got about 150 hosts in it > total, and the screen we've designed for our app (which has about 15 > graphs on it, not terribly much), takes SO LONG to generate the graph > images that -- heh, wait for it -- the Refresh-delay kicks it, and the > browser ceases to care about those graphs it has spent so much time > generating and has asked for a WHOLE NEW set of graphs. This is because it doesn't use RRDs, it keeps each data point in the database until it expires (default 1 year) which will create a huge data set. Pretty stupid design decision IMHO. > -- You want to do some sort of programatic creation of hosts/services/ > etc. OTHER THAN via the auto-discovery method? Ha! Good luck with > that. Doesn't seem to be any good means of doing so. Programaticly, not exactly. But you can create groups and assign templates to those groups. Unfortunately this isn't intuitive and last time I tried it I gave up. > -- Zabbix's "all or nothing" definition of "monitoring" is, for our > uses, completely useless. There's no concept of warnings versus > critical, or perhaps there's times when that check isn't as important > as it might be at other times, and no way to squelch the messages or > anything like that. Trending, you can at least claim is Zabbix's > specialty, but we've just completely ripping its "monitoring and > notification" aspects completely out and doing that in parallel with > Nagios. Triggers can be any one of five severities, and you can trigger on any level of any data item. So, for instance, you can have a warning when drive space is down to 10% and a critical alert when it is down to 2%. But if you are creating this many alerts you better be using templates, so they they get automatically added to everything using that template. You can also specify that a data item is only checked during certain times of the day. This is all a very complex way of creating different warning levels at different times of the day, but it /can/ do that. You can also create notifications that send different alert levels to different people at different times of the day. -- Perfection is just a word I use occasionally with mustard. --Atom Powers-- _______________________________________________ Discuss mailing list [email protected] http://lopsa.org/cgi-bin/mailman/listinfo/discuss This list provided by the League of Professional System Administrators http://lopsa.org/
