On 30/03/15 12:52, Craig Small wrote:
> On Mon, Mar 30, 2015 at 11:43:21AM +0100, Vsevolod Stakhov wrote:
>> Could you please paste strace of rspamd worker when the timeout occurs?
> epoll_wait(10, {{EPOLLIN, {u32=3, u64=3}}}, 32, 20267) = 1
> accept(3, {sa_family=AF_INET6, sin6_port=htons(38141), inet_pton(AF_INET6,
> "::ffff:127.0.0.1", &sin6_addr), sin6_flowinfo=0, sin6_scope_id=0}, [28]) = 21
> fcntl(21, F_GETFL) = 0x2 (flags O_RDWR)
> fcntl(21, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> fcntl(21, F_SETFD, FD_CLOEXEC) = 0
> clock_gettime(CLOCK_PROCESS_CPUTIME_ID, {3, 396313080}) = 0
> stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=2197, ...}) = 0
> writev(2, [{"2015-03-30 22:37:07 #32545(norma"..., 50}, {"accepted connection
> from ::ffff:"..., 52}, {"\n", 1}], 3) = 103
> epoll_ctl(10, EPOLL_CTL_ADD, 21, {EPOLLIN, {u32=21, u64=21}}) = 0
> epoll_wait(10, {{EPOLLIN, {u32=21, u64=21}}}, 32, 12628) = 1
> epoll_ctl(10, EPOLL_CTL_DEL, 21, 7fffac3ee930) = 0
> stat("/etc/rspamd/2tld.inc", {st_mode=S_IFREG|0644, st_size=86920, ...}) = 0
> epoll_ctl(10, EPOLL_CTL_ADD, 21, {EPOLLIN, {u32=21, u64=21}}) = 0
> stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=2197, ...}) = 0
> writev(2, [{"2015-03-30 22:37:07 #32545(norma"..., 64}, {"abnormally closing
> connection fr"..., 71}, {"\n", 1}], 3) = 136
> close(21) = 0
>
> I am using IPv6 here, so possibly something inside rspam isn't liking
> it. You can see in the accept function the IP address.
>
This is very weird behaviour. As you can see, epoll_wait returns 1 but
there is no attempt to read anything from a socket, however, according
to the code this could happen if and only if libevent returns EV_TIMEOUT:
https://github.com/vstakhov/rspamd/blob/rspamd-0.8/src/libutil/http.c#L668
Just curious, does this problem exist if you edit your workers.conf and
set parameter `count` to 1 for normal worker there?
--
Vsevolod Stakhov
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
