docker ps hangs #7905
Description
When running docker 1.2.0, after a variable amount of time (~0.5h typically), docker starts hanging on certain tasks.
docker ps in particular does not complete after this issue appears. Here's an strace of the server and client sides:
Client
I waited about 45 seconds before killing the straces.
write(3, "GET /v1.14/containers/json HTTP/"..., 123) = 123
epoll_wait(4, {{EPOLLOUT, {u32=1591284800, u64=140038999970880}}}, 128, 0) = 1
epoll_wait(4, {{EPOLLOUT, {u32=1591284800, u64=140038999970880}}}, 128, -1) = 1
epoll_wait(4, ^CProcess 8835 detached
Server
There were no further mentions of FD 206 after this excerpt.
[pid 10608] 16:45:54.500386 <... accept4 resumed> {sa_family=AF_LOCAL, NULL}, [2], SOCK_CLOEXEC|SOCK_NONBLOCK) = 206 <0.000013>
[pid 10608] 16:45:54.500407 epoll_ctl(7, EPOLL_CTL_ADD, 206, {EPOLLIN|EPOLLOUT|EPOLLRDHUP|EPOLLET, {u32=940253712, u64=140064118743568}}) = 0 <0.000010>
[pid 10608] 16:45:54.500437 getsockname(206, {sa_family=AF_LOCAL, sun_path="/var/run/docker.sock"}, [23]) = 0 <0.000007>
[pid 10579] 16:45:54.500468 <... select resumed> ) = 0 (Timeout) <0.000086>
[pid 10608] 16:45:54.500476 accept4(6, <unfinished ...>
[pid 10579] 16:45:54.500484 select(0, NULL, NULL, NULL, {0, 20} <unfinished ...>
[pid 10608] 16:45:54.500493 <... accept4 resumed> 0xc2080b97a8, [112], SOCK_CLOEXEC|SOCK_NONBLOCK) = -1 EAGAIN (Resource temporarily unavailable) <0.000012>
[pid 10608] 16:45:54.500520 read(206, "GET /v1.14/containers/json HTTP/"..., 4096) = 123 <0.000011>
[pid 10579] 16:45:54.500570 <... select resumed> ) = 0 (Timeout) <0.000080>
[pid 10579] 16:45:54.500582 select(0, NULL, NULL, NULL, {0, 20} <unfinished ...>
[pid 10608] 16:45:54.500645 write(1, "[info] GET /v1.14/containers/jso"..., 34) = 34 <0.000013>
We encounter this problem with docker 1.2.0, but not docker 1.0.0, on ubuntu:14.04 with both 3.13 and 3.16 kernels.