[Pdns-users] pdns-recursor cache size on 3.3 versus 3.1
Simon Kirby
sim at hostway.ca
Mon Aug 27 23:19:39 UTC 2012
Hello!
Since upgrading pdns-recursor instances in a mail cluster environment
from 3.1 to 3.3, we noticed that the positive and negative cache sizes
seem to grow much larger, and don't seem to follow the same pattern they
used to, which seemed to look more like what one might expect with
average encountered TTLs. Since the upgrade, I have never seen the cache
shrink, only grow.
I posted these graphs earlier, showing the situation about two weeks
after the upgrade: http://0x.ca/sim/tmp/nsc08-pdns_recursor-upgrade.html
It started hitting swap at our original max-cache-entries setting (9M),
so we've restarted it with 6M, and it's just about to reach that now:
[sroot at nsc08:/root]# rec_control get-all | awk '{printf "%10u %s\n",$2,$1}'
30227604 all-outqueries
0 dlg-only-drops
1961 dont-outqueries
678286 outgoing-timeouts
40871 tcp-outqueries
185278 throttled-out
185278 throttled-outqueries
43632 unreachables
531886 answers-slow
10631721 answers0-1
1526028 answers1-10
10980909 answers10-100
4789382 answers100-1000
0 case-mismatches
247107 chain-resends
0 client-parse-errors
0 edns-ping-matches
0 edns-ping-mismatches
0 ipv6-outqueries
28218834 no-packet-error
30268556 noedns-outqueries
186307287 noerror-answers
0 noping-outqueries
25088 nsset-invalidations
30361670 nxdomain-answers
0 over-capacity-drops
0 qa-latency
217550214 questions
0 resource-limits
590 server-parse-errors
881109 servfail-answers
0 spoof-prevents
0 tcp-client-overflow
8985 tcp-questions
0 unauthorized-tcp
0 unauthorized-udp
878 unexpected-packets
5994112 cache-entries
9688816 cache-hits
18771115 cache-misses
3 concurrent-queries
600006 negcache-entries
7076 nsspeeds-entries
500032 packetcache-entries
189090331 packetcache-hits
28450967 packetcache-misses
3683842 sys-msec
0 tcp-clients
143 throttle-entries
603558 uptime
11785408 user-msec
What were throttled outqueries seem to have been replaced with
outqueries. This probably resulted in the latency changes. I suspect
the packet cache is probably causing the lower main cache hit rate.
Simon-
More information about the Pdns-users
mailing list