Bug#540775: nscd: using 100% of cpu immediately after start
On Mon, Aug 10, 2009 at 10:28:41AM +0000, Siim Põder wrote:
> Package: nscd
> Version: 2.7-18
> Severity: important
>
>
> Immediately after starting nscd it starts using 100% cpu, strace showing:
>
> [pid 15848] futex(0x552acc75e0, FUTEX_WAKE, 1 <unfinished ...>
> [pid 15845] write(2, "15842: short read while reading r"..., 7415842: short read while reading request: Resource temporarily
> unavailable
> <unfinished ...>
> [pid 15847] close(11 <unfinished ...>
> [pid 15842] epoll_wait(10, <unfinished ...>
> [pid 15848] <... futex resumed> ) = 0
> [pid 15845] <... write resumed> ) = 74
> [pid 15847] <... close resumed> ) = 0
> [pid 15842] <... epoll_wait resumed> {}, 100, 4292900169) = 0
> [pid 15848] futex(0x552acc7540, FUTEX_WAKE, 1 <unfinished ...>
> [pid 15845] close(11 <unfinished ...>
> [pid 15847] futex(0x552acc75e4, FUTEX_WAIT, 18, NULL <unfinished ...>
> [pid 15842] epoll_wait(10, <unfinished ...>
> [pid 15848] <... futex resumed> ) = 0
> [pid 15845] <... close resumed> ) = -1 EBADF (Bad file descriptor)
> [pid 15842] <... epoll_wait resumed> {{EPOLLRDNORM, {u32=9, u64=9}}}, 100, 4292900169) = 1
> [pid 15848] fcntl(11, F_GETFL <unfinished ...>
> [pid 15845] clock_gettime(CLOCK_MONOTONIC, <unfinished ...>
> [pid 15842] accept(9, <unfinished ...>
> [pid 15848] <... fcntl resumed> ) = -1 EBADF (Bad file descriptor)
> [pid 15845] <... clock_gettime resumed> {37414733, 947065224}) = 0
> [pid 15842] <... accept resumed> 0, NULL) = 11
> [pid 15848] close(11 <unfinished ...>
> [pid 15845] futex(0x552acc75e4, FUTEX_WAIT, 19, {0, 168829000} <unfinished ...>
> [pid 15842] epoll_ctl(10, EPOLL_CTL_ADD, 11, {EPOLLRDNORM, {u32=11, u64=11}} <unfinished ...>
> [pid 15848] <... close resumed> ) = 0
> [pid 15842] <... epoll_ctl resumed> ) = -1 EBADF (Bad file descriptor)
> [pid 15848] futex(0x552acc75e4, FUTEX_WAIT, 20, NULL <unfinished ...>
> [pid 15842] close(11) = -1 EBADF (Bad file descriptor)
> [pid 15842] epoll_wait(10, {}, 100, 4292900169) = 0
> [pid 15842] epoll_wait(10, {}, 100, 4292900169) = 0
> [pid 15842] epoll_wait(10, {}, 100, 4292900169) = 0
> [pid 15842] epoll_wait(10, {}, 100, 4292900169) = 0
> [pid 15842] epoll_wait(10, {}, 100, 4292900169) = 0
> ....
>
> epoll_wait ad nauseam
>
> This may or may not be the same bug as #513635:
> * the system is also using ldap
> * there is just one nscd thread
> * number of fds used is low while there is 100% CPU usage
> * nscd seems to work find in spite of this
>
>
> -- System Information:
> Debian Release: 5.0.2
> APT prefers stable
> APT policy: (500, 'stable')
> Architecture: amd64 (x86_64)
>
> Kernel: Linux 2.6.8-12-amd64-k8-smp (SMP w/2 CPU cores)
> Locale: LANG=en_US, LC_CTYPE=en_US (charmap=ISO-8859-1)
> Shell: /bin/sh linked to /bin/bash
>
The bad descriptors looks very strange. Have you tried with a more
recent kernel? Stable ships with a 2.6.26 kernel.
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurelien@aurel32.net http://www.aurel32.net
Reply to: