DNS test timeout for a downed DNS server.
list Steve Holmes
Xymon 4.2.3. One of the DNS servers we monitor was down. The bbtest total time went from under 30 seconds to over 900 seconds almost all of which was for the DNS test. This caused a number of problems. I removed the dns tag from the down server and the time dropped back to the normal level. The mysteries to me are why did Xymon continue to test the DNS service on that server while the conn test was red? Why did it take 900 seconds for it to time out, even though I was running with --dns-timeout=60? ( I.e. I can understand that it might take longer than with the default dns timeout, but not 900 seconds.) The color for the dns test for that host was clear and displayed the following text: Service dns on dns.server.org is OK Dialup host/service, or test depends on another failed test Host appears to be down Timeout Seconds: 900.042 It also stopped recording data in the graph. And more importantly, is the behavior still the same in 4.3.9? Thanks, Steve Holmes Purdue University
list Steven Carr
Check the archives, there is a well documented issue with the DNS timeout not working as you think it should work (to do with the underlying libraries). IIRC some fixes/workarounds have been added to later versions but it wont be til the next major release (5.?) when the underlying testing tool has been rewritten that it will be fixed for good. Steve
▸
On 24 July 2012 21:17, Steve Holmes <user-ec1bf77b1b44@xymon.invalid> wrote:
Xymon 4.2.3. One of the DNS servers we monitor was down. The bbtest total time went from under 30 seconds to over 900 seconds almost all of which was for the DNS test. This caused a number of problems. I removed the dns tag from the down server and the time dropped back to the normal level. The mysteries to me are why did Xymon continue to test the DNS service on that server while the conn test was red? Why did it take 900 seconds for it to time out, even though I was running with --dns-timeout=60? ( I.e. I can understand that it might take longer than with the default dns timeout, but not 900 seconds.) The color for the dns test for that host was clear and displayed the following text: Service dns on dns.server.org is OK Dialup host/service, or test depends on another failed test Host appears to be down Timeout Seconds: 900.042 It also stopped recording data in the graph. And more importantly, is the behavior still the same in 4.3.9? Thanks, Steve Holmes Purdue University
list Henrik Størner
▸
On 24-07-2012 22:17, Steve Holmes wrote:
Xymon 4.2.3. One of the DNS servers we monitor was down. The bbtest total time went from under 30 seconds to over 900 seconds almost all of which was for the DNS test. This caused a number of problems. I removed the dns tag from the down server and the time dropped back to the normal level. The mysteries to me are why did Xymon continue to test the DNS service on that server while the conn test was red? Why did it take 900 seconds for it to time out, even though I was running with --dns-timeout=60?
DNS timeout handling does not work correctly in versions prior to 4.3.8, it is a case of the DNS library (c-ares) not behaving the way I thought it did. 4.3.8+ has a workaround in place so the timeout is fixed at approximately 25 seconds. You can achieve almost the same result by running the old bbtest-net utility with "--dns-timeout=2" (no, this won't cause DNS lookups to timeout after 2 seconds).
And more importantly, is the behavior still the same in 4.3.9?
No. Regards, Henrik