Xymon Mailing List Archive search

Hobbit Sensitivity lowered?

5 messages in this thread

list Chris Wopat · Wed, 08 Feb 2006 11:20:30 -0600 ·
Hello,

I switched from BB to hobbit about a month ago and have been generally happy. One thing that's "different" is that Hobbit seems to be more sensitive to services timing out, such as during an http check.

I don't see anything in bb-services related to this, is there any global setting somwhere to permit me to allow a check to take a few more ms to respond before it's considered "red"? My bb2.html file is frequently red due to this, although it only lasts a few seconds.

Is there an appropriate way to do this without hacking code and finding the timeout period?

Thanks,
--Chris
list Henrik Størner · Wed, 8 Feb 2006 21:39:37 +0100 ·
quoted from Chris Wopat
On Wed, Feb 08, 2006 at 11:20:30AM -0600, Chris Wopat wrote:
I switched from BB to hobbit about a month ago and have been generally happy. One thing that's "different" is that Hobbit seems to be more sensitive to services timing out, such as during an http check.
The default timeout for network tests in Hobbit is 10 seconds. You can
change it via the --timeout=N option (N is in seconds) for bbtest-net in hobbitlaunch.cfg.  See the bbtest-net(1) man-page...

BB uses a fixed 20 second timeout.
I don't see anything in bb-services related to this, is there any global setting somwhere to permit me to allow a check to take a few more ms to respond before it's considered "red"?
Millisecs shouldn't make any difference. It's a global setting - not in
the bb-services file per service - because all of the network tests
really run in parallel. Having individual timeouts is just
un-manageable.
My bb2.html file is frequently red due to this, although it only lasts a few seconds.
How many tests do you run ? (See the status in the "bbnet" column for
your Hobbit server). It's a well-known problem that Hobbit can actually
flood some networks/firewalls with connection setups, resulting in some tests showing up as failed one moment, and OK a few seconds after.
You might want to lower the number of concurrent tests that bbtest-net
runs via the --concurrency=N option - the default is 256.
Is there an appropriate way to do this without hacking code and finding the timeout period?
Read The Fine Manual :-) You'll find them on your Hobbit server webpages, in the "Help"->"Hobbit man-pages" section.


Regards,
Henrik
list Chris Wopat · Wed, 08 Feb 2006 16:55:12 -0600 ·
quoted from Henrik Størner
Henrik Stoerner wrote:
On Wed, Feb 08, 2006 at 11:20:30AM -0600, Chris Wopat wrote:
I switched from BB to hobbit about a month ago and have been generally happy. One thing that's "different" is that Hobbit seems to be more sensitive to services timing out, such as during an http check.

The default timeout for network tests in Hobbit is 10 seconds. You can
change it via the --timeout=N option (N is in seconds) for bbtest-net in hobbitlaunch.cfg.  See the bbtest-net(1) man-page...

BB uses a fixed 20 second timeout.
Thanks, I will bump this to, say, 15 and test. I was misreading the graphs in the hobbit pages at times to be ms, when really it's seconds.
quoted from Henrik Størner

How many tests do you run ? (See the status in the "bbnet" column for
your Hobbit server). It's a well-known problem that Hobbit can actually
flood some networks/firewalls with connection setups, resulting in some tests showing up as failed one moment, and OK a few seconds after.
You might want to lower the number of concurrent tests that bbtest-net
runs via the --concurrency=N option - the default is 256.
According to bb-hosts, it looks like 290 hosts, with an average of 2-3 tests per (many are just single pings, about 80 have http checks and a few others). I'm unsure what you speak of when you say "bbnet column"- in the html result pages, or in server/etc/* ?
quoted from Henrik Størner
Is there an appropriate way to do this without hacking code and finding the timeout period?

Read The Fine Manual :-) You'll find them on your Hobbit server webpages, in the "Help"->"Hobbit man-pages" section.
Aye. I'll try the first timeout to be a bit longer, then tweak the concurrent tests if need be.

Thanks,
--Chris
list Olivier Beau · Fri, 10 Feb 2006 15:58:07 +0100 ·
maybe for version 12 of hobbitd.. ;)
http://www.synderella.org/


olivier
list Henrik Størner · Fri, 10 Feb 2006 16:56:13 +0100 ·
On Fri, Feb 10, 2006 at 03:58:07PM +0100, Olivier Beau wrote:
maybe for version 12 of hobbitd.. ;)
http://www.synderella.org/
Amazing what people can come up with...

I think that should be for version 8. By the time we get to
version 12, Hobbit will be streaming alerts directly into
your cerebral cortex (we'll all have wi-fi implants done at
birth).


Regards,
Henrik