[Pdns-users] High CPU load with no traffic

[Ciro Iriarte]

Hi!, I'm deplaying some recursors with 3.6.2 (packaged by EPEL) and
they randomly use a lot of CPU and the queries dy by timeout. I've
seen loads from 1 to 5 (100% and 500% CPU usage).

This happened with 3 out of 7 servers, a restart temporaly fixed the
issue, one of them had the same behaviour within minutes, the others
are still fine. tcpdump shows not traffic.

Anybody has seen this?

Configuration:

allow-from-file = /etc/pdns-recursor/clients.acl
any-to-tcp = yes
local-address = 127.0.0.1,X.X.X.X, X.X.X.X, X.X.X.X
lua-dns-script = /etc/pdns-recursor/my.lua
pdns-distributes-queries = yes
setgid = pdns-recursor
setuid = pdns-recursor
threads = 16
version-string = XXXXX

Package:

pdns-recursor-3.6.2-1.el6.x86_64

strace doesn't show much:

---
[pid  6050] epoll_wait(95,  <unfinished ...>
[pid  6049] epoll_wait(86,  <unfinished ...>
[pid  6048] epoll_wait(91,  <unfinished ...>
[pid  6047] epoll_wait(93,  <unfinished ...>
[pid  6046] epoll_wait(85,  <unfinished ...>
[pid  6045] epoll_wait(79,  <unfinished ...>
[pid  6044] epoll_wait(94,  <unfinished ...>
[pid  6043] epoll_wait(88,  <unfinished ...>
[pid  6041] epoll_wait(82,  <unfinished ...>
[pid  6040] epoll_wait(81,  <unfinished ...>
[pid  6039] epoll_wait(83,  <unfinished ...>
[pid  6038] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6037] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6035] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6038] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6037] <... rt_sigprocmask resumed> [], 8) = 0
[pid  6036] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6035] <... rt_sigprocmask resumed> [], 8) = 0
[pid  6038] <... rt_sigprocmask resumed> [], 8) = 0
[pid  6037] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6036] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6038] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6037] <... rt_sigprocmask resumed> NULL, 8) = 0
[pid  6035] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6036] <... rt_sigprocmask resumed> [], 8) = 0
[pid  6038] <... rt_sigprocmask resumed> NULL, 8) = 0
[pid  6037] epoll_wait(90,  <unfinished ...>
[pid  6035] <... rt_sigprocmask resumed> NULL, 8) = 0
[pid  6038] epoll_wait(92,  <unfinished ...>
[pid  6036] rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
[pid  6035] epoll_wait(80,  <unfinished ...>
[pid  6036] <... rt_sigprocmask resumed> NULL, 8) = 0
[pid  6036] epoll_wait(87,  <unfinished ...>
[pid  6050] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6049] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6048] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6047] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6046] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6045] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6044] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6043] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6041] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6040] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6039] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6050] epoll_wait(95,  <unfinished ...>
[pid  6049] epoll_wait(86,  <unfinished ...>
[pid  6048] epoll_wait(91,  <unfinished ...>
[pid  6047] epoll_wait(93,  <unfinished ...>
[pid  6046] epoll_wait(85,  <unfinished ...>
[pid  6045] epoll_wait(79,  <unfinished ...>
[pid  6044] epoll_wait(94,  <unfinished ...>
[pid  6043] epoll_wait(88,  <unfinished ...>
[pid  6041] epoll_wait(82,  <unfinished ...>
[pid  6040] epoll_wait(81,  <unfinished ...>
[pid  6039] epoll_wait(83,  <unfinished ...>
[pid  6038] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6037] <... epoll_wait resumed> {}, 1024, 500) = 0
[pid  6035] <... epoll_wait resumed> {}, 1024, 500) = 0
---

Regards,

-- 
Ciro Iriarte
http://iriarte.it
--