Quack.id.au

Needed to get Zenoss alerting sensably for SNMP Traps generated by VMWare VC… here is the Event Transform that I have come up with:

# Data from the SNMP Trap
#vmwVpxdTrapType = getattr(evt, "vmwProductSpecific.3.301.0", "Unknown")
#vmwVpxdHostName = getattr(evt, "vmwProductSpecific.3.302.0", "Unknown")
#vmwVpxdVMName = getattr(evt, "vmwProductSpecific.3.303.0", "Unknown")
#vmwVpxdOldStatus = getattr(evt, "vmwProductSpecific.3.304.0", "Unknown")
vmwVpxdNewStatus = getattr(evt, "vmwProductSpecific.3.305.0", "Unknown")
vmwVpxdObjValue = getattr(evt, "vmwProductSpecific.3.306.0", "Unknown")
vmwVpxdTargetObj = getattr(evt, "vmwProductSpecific.3.307.0", "Unknown")
#vmwVpxdTargetObjType = getattr(evt, "vmwProductSpecific.3.308.0", "Unknown")
# Severity
if vmwVpxdNewStatus == "Red":
	evt.severity = 5
elif vmwVpxdNewStatus == "Yellow":
	evt.severity = 4
elif vmwVpxdNewStatus == "Green" or vmwVpxdNewStatus == "Gray":
	evt.severity = 0
else:
	evt.severity = 3
# Component
evt.component = vmwVpxdTargetObj
# Summary
evt.summary = "Trap " + vmwVpxdTargetObj
# Message
evt.message = vmwVpxdTargetObj + ": " + vmwVpxdObjValue

At work we are hanging on to our voice E1 PRI as, we figure, as a networking company, we don’t want a network issue to affect the only non-computer-based way for customers to get in touch with us (Phone). The other reason is because, for some reason, the Integrated Public Number Database (IPND), still requires all updates to be submitted via an ISDN dial-up connection. The IPND’s primary function is to hold a list of phone numbers and addresses for use by the emergency services to find your house if you dial 000.

We very recently moved the termination of this E1/PRI to a Digium card in a box running Asterisk, previously this was done on a Cisco 5300. After a bit of reading it became apparent that someone had already built all that was required to dial out via a Digium PCI card. Oh JOY!

The basics of getting this to work are:

1. Ensure your kernel is compiled with support for Generic HDLC interfaces (Device Drivers > Network Interfaces > WAN Interfaces) either statically/module. I compiled it all as a module, per the CentOS guide.
2. If PPP is not compled statically into kernel you can also have problems with DAHDI recognising that it’s supported hence will not compile its own support for PPP. I forced PPP support by editing include/dahdi/dahdi_config.h
3. Told asterisk (statically) to not use Channel 31 on the PRI (editing DAHDI module config for Asterisk)
4. Created a peer config /etc/ppp/peers/ipnd:

   lock
   dump
   debug
   logfile /var/log/pppd
   plugin dahdi.so
   31
   noauth
   user test
   connect “chat -t 3 -f /etc/ppp/peers/ipnd_chat”

5. Will also have to create the chat script and verify all that works, once we have details from the IPND people. But so far, so good.

References:

• http://wiki.centos.org/HowTos/BuildingKernelModules
• http://www.facebook.com/note.php?note_id=440833568415

For a couple of months free memory has been on a downward spiral, and CPU utilisation steadily rising. After much consultation, confusion and head scratching we came across a weird and wacky bug with our Cisco 7200 series routers (possibly just the firmware we are running)…

Having BGP peers that are shutdown or otherwise inactive will cause the BGP process to gradually chew through available memory, and also cause it to use excessive CPU resources. Removing the BGP configuration for the inactive peers will release the held memory and drop CPU utilisation significantly. Our routers immediately recovered 60-100MB of memory and is now on an upward trend, and CPU utilisation dropped by around 50 percent.

Have observed the phenomenon on

• Cisco 7206 VXR NPE-G1 running c7200-advipservicesk9-mz.124-24.T.bin
• Cisco 7206 VXR NPE-G1 running c7200-spservicesk9-mz.124-24.T.bin