On 3/12/10 12:26 PM, Pat Storr wrote: > We're polling more than 500 devices, for SNMP or ICMP, generally at 30 > second intervals. Intermittently, IM will report a random device(s) as being > "down". A console ping session from the IM server confirms the IM > status...that the device is unavailable. Traceroute from the IM server stops > ONE HOP AWAY from the device reported as down. During the time that IM > reports the device to be down we can verify that the device is indeed alive > by pinging from another PC. The device may stay "down" anywhere from a few > seconds to several minutes...long enough to set off alerts to our > pagers/cell phones. > > We can clean up the maps by simply restarting the IM process on the server. > A ping session running from the IM server while we restart the process > confirms the result. > We've received similar reports in the past. Machine S running InterMapper suddenly stops being able to ping machine X. This loss of connectivity with machine X is confirmed with other ping tools on the same machine S. Machine S has no problem pinging another machine (X+1) on the same subnet as X. In past reports, rebooting S fixes the problem for a while...
My first suspicion is that machine S has a bad route to machine X. Run 'netstat -ran' on machine S to see the routing table. Run 'netstat -rs' to see the routing stats. Running 'netstat -s' may be enlightening if you look at the histogram of ICMP types received and there are suspicious ICMP redirects present. I am puzzled that restarting the IM process on the server (without rebooting the server or toggling the network connection) can cause connectivity to recover. The only possible change is that IM is closing and re-opening the raw IP socket used for pings, and UDP-based probes may also obtain new source port numbers. IM does not touch the local routing table, nor does InterMapper do anything to change the network status. I am very interested in collecting information from other people who are seeing this behavior. Thanks. -- Bill Fisher Dartware, LLC ____________________________________________________________________ List archives: http://www.mail-archive.com/intermapper-talk%40list.dartware.com/ To unsubscribe: send email to: [email protected]
