On 05/23/2013 02:20 PM, Wenhao wrote: > On Thu, May 23, 2013 at 2:01 PM, Liu Yuan <namei.unix at gmail.com> wrote: >> Signed-off-by: Liu Yuan <namei.unix at gmail.com> >> --- >> sheep/gateway.c | 4 +++- >> 1 file changed, 3 insertions(+), 1 deletion(-) >> >> diff --git a/sheep/gateway.c b/sheep/gateway.c >> index 1a37ad8..d177345 100644 >> --- a/sheep/gateway.c >> +++ b/sheep/gateway.c >> @@ -164,7 +164,9 @@ again: >> >> panic("%m"); >> } else if (pollret == 0) { >> - sd_eprintf("poll timeout %d", wi->nr_sent); >> + sd_eprintf("poll timeout %d, the disks of some nodes or " >> + "network is busy. Going to poll-wait again", >> + wi->nr_sent); >> /* >> * If IO NIC is down, epoch isn't incremented, so we can't retry >> * for ever. > > This seems not a good place to print "Going to poll-wait again". I > guess it should be placed just before > if (sheep_need_retry(req->rq.epoch) && repeat) { > repeat--; > >>> goto again; Ah yeah, well. if not poll again, the request will be retried by gateway, see gateway_io_done(), so else isn't useful, user are not interested in it. We print poll timeout just as a suggestion that disks or network is busy, user might enlarge network bandwidth or add more nodes to share disk IOs. Thanks, Yuan |