Servicemon/SmbChecker runs out of available file descriptors

[Morten Brekkevold]

A bug in the servicemon checker plugin SmbChecker causes the servicemon daemon to keep opening new UNIX pipes, until it hits the system limit on the number of open files. This may also affect the rest of the processes running as the navcron user, as the limit is often per-user.

This bug only appears when an smb service handler has been added.

[Morten Brekkevold (mbrekkevold)]

I propose the attached patch as a fix for this, but I will need to test it some more first.

[Morten Brekkevold (mbrekkevold)]

I have investigated this further. This actually consists of two issues. The minor issue is the typo in the SmbChecker code, but is actually not this typo that is causing the number of open pipes to increase indefinitely.

This seems to be a known bug in Python, see http://bugs.python.org/issue5179 . Unfortunately for us, it has only been fixed in Python 2.6.

IOW, this bug will only affect you if you are running Python < 2.6 and enable the Samba service checker without having smbclient installed. Apart from fixing the typo, it seems the best we can do is write a workaround for the Python bug - something akin to verifying that there is actually an executable smbclient present before attempting to run it.

[Morten Brekkevold (mbrekkevold)]

Re-assigning this to Ole Martin, since he's the one mainting the service monitor nowadays :)

[Morten Brekkevold (mbrekkevold)]

Re-assigning this to Ole Martin, since he's the one maintaining the service monitor nowadays :)

[Ole Martin Bjørndalen (olemb)]

I've written a find_cmd() for the new DHCP checker which can be moved to a library and used to fix this bug:

https://heimdal.uit.no/hg/olemb/dhcpchecker/file/tip/subsystem/statemon/nav/statemon/checker/DhcpChecker.py

[Ole Martin Bjørndalen (olemb)]

Just not sure where to put it.

[Morten Brekkevold (mbrekkevold)]

Ole Martin, I am unable to find your bugfix for the 3.5.x branch? The link to your repository gives my a 500 internal server error...

[Kai Bjørnenak (kai-bjornenak)]

Fix can be found here: https://heimdal.uit.no/hg/olemb/servicemon-spring-cleaning/rev/e3fabef80e8b

[Morten Brekkevold (mbrekkevold)]

This fix seems to depend on a whole range of changes Ole Martin has made to clean the statemon code. I'm therefore hesitant to make this part of NAV 3.7. I've merged the branch to the default branch, and am therefore retargeting this to NAV 3.8.0.

Merged changeset here: http://metanav.uninett.no/hg/default/rev/e3fabef80e8b