Nagios + 6.3-RELEASE == Hung Process
jarrod at netleader.com.au
Wed Jan 2 16:35:24 PST 2008
On Wed, 2 Jan 2008, Tom Judge wrote:
> Jarrod Sayers wrote:
>> I hope I can confirm your frustrations. There is a threading issue
>> with Nagios when it's binaries are linked against libpthread(3)
>> threading library, the default on recent FreeBSD 5.x releases and all
>> 6.x releases. The issue is random and extremely difficult to track down
>> with the symptoms being a second Nagios process sitting on the system
>> hanging a CPU. Be rest assured that I have been working on it, and
>> have seen it on one system of mine.
> Not sure if this is related at all but out of the 3 nagios deployments
> we have here I have only ever seen it on one (It currently has 2 nagios
> threads spinning CPU time atm).
> The differences on that server are:
> * It is amd64 compared to i386
> * It also runs ndo2db from ndoutils 1.4b7
> All the systems run 6.2-RELEASE-p5 and nagios-2.9_1, they are also all
> patched with gnu libltdl patch below.
> Don't know if that info is of any use to you.
That's actually good to know, as you're now (unless I am mistaken) the
first user to contact me about this problem on non-i386 systems. One
user, plus myself, have also seen the issue under Nagios 3.x, both on i386
I also have a net-mgmt/ndoutils port in the works (less the database
support for now) which also has the same issue so using broker modules
doesn't seem to affect the outcome.
My gut feeling is that it's not an architecture issue but more an
interoperability issue between the Nagios threading code and the
libpthread() threading library.
>> I did receive that email and the changes went in with the last commit
>> of net-mgmt/nagios-devel to test. No issues have arisen so i'll be
>> back-porting it to net-mgmt/nagios soon for you. There also has been a
>> rather large ports freeze which delayed the upgrade to Nagios 2.10,
>> that PR was submitted on the 1st of November and committed on the 13th
>> of December. Unfortunately your email fell somewhere in the middle,
>> apologies for not letting you know.
> Thanks for this, I currently maintain the patch on our build servers.
No worries, I will look at bundling in the change with the libthr() fix
over the next few days. Thanks for pointing that out too as it was a bug
instead of a feature request, as on systems where the library was
available, the build process would link to it. Hmm...
More information about the freebsd-stable