Hello,
I am wondering if anyone has any ideas on how to help me with DJB’s DNSCACHE program. Though I am wondering if the issue is DJB or the OS. Running Redhat 5.3 to 5.8 at the moment.
We recently had a DNS spike. I found the issue to be the cache size was set way to low, and we seemed to be peaking at 1020ish currnet requests per server. ( we have 4 ) When I increased the cache size it was able serve more queries faster and the concurrent number dropped. Back to around 300 or so like it usually is. This got me looking though, into why it wouldn’t go over 1020ish queries. I know we have gone above that in the past. Our servers are compiled at MAXUDP 3000. Our Run script has the softlimit set to 2500. When our machines were not as strong as they are today we used to get up to around 2100 queries per box. I wrote a little program that would monitor the logs and send me an email when our queries reached 2400 so I knew something was up. This happen on several occasions so I know it used to work.
A couple years ago we moved to new data centers and built new machines. A couple of the settings were not set right hence the issue with the cache listed above. The issue I am facing now is I can’t get my current queries to break the 1024 mark. I did some digging and it was suggested that this might be an issue of file descriptors. I modified the system so the device limit is set at 500,000 and the user hard and soft limits are set 60,000. Still I can’t get dnscache to go over 1024.
For performance testing I am using a program called resperf ( part of dnsperf ) The file I am using has 10 million domain entries in it. When I run this against my test server I can make it hit the 1024 cap in about 5 seconds and keep it there but not over, until the perf program tells me there are 65000+ queries waiting. This app it run from another server.
Any ideas what could be causing this limit?
Regards,