[ovs-dev] [PATCHv3] monitor: Replace monitor_seq with periodic wakeup

Alex Wang alexw at nicira.com
Fri Dec 13 05:50:34 UTC 2013


Yes, I could reproduce it on my VM setup,

It is the timing issue again, if the p1 starts sending ccm first, it will
have flap count = 0.

If p1 receives ccm from p0 first, it will go to [rdi] and then [].  so it
will have flap count = 2.

I think the safest way is to drop the "CFM_VSCTL_LIST_IFACE([p1],
[cfm_flap_count], [cfm_flap_count      : 0])" and don't check
p1 flap count.




On Thu, Dec 12, 2013 at 8:20 PM, Ethan Jackson <ethan at nicira.com> wrote:

> This patch occasionally fails the cfm flap count test for me.  Can
> either of you reproduce it?
>
> Ethan
>
> On Thu, Dec 12, 2013 at 5:52 PM, Joe Stringer <joestringer at nicira.com>
> wrote:
> > Currently, every time a monitoring port is added or reconfigured, the
> > main thread notifies the monitoring thread to wake up immediately using
> > monitor_seq. When adding a large number of ports at once, this causes
> > contention as the threads fight over access to the monitor heap---one
> > thread adding new ports, the other popping and processing the new ports.
> >
> > This patch removes this mechanism in favour of a simple periodic wakeup
> > in the monitor thread. When configuring batches of 500 tunnels at once,
> > up to a total of 5000 tunnels, this reduces configuration time by up to
> > 35%. New port monitor port configuration may not take effect for as long
> > as ~100ms, although it may be less if the timer has not recently timed
> > out, or if a monitoring port is due to send a packet.
> >
> > Signed-off-by: Joe Stringer <joestringer at nicira.com>
> > ---
> > v3: Rebase
> > ---
> >  ofproto/ofproto-dpif-monitor.c |   18 ++++++++++--------
> >  tests/bfd.at                   |   16 +++++++---------
> >  tests/cfm.at                   |    8 +++++---
> >  3 files changed, 22 insertions(+), 20 deletions(-)
> >
> > diff --git a/ofproto/ofproto-dpif-monitor.c
> b/ofproto/ofproto-dpif-monitor.c
> > index d06b2e1..2a833bb 100644
> > --- a/ofproto/ofproto-dpif-monitor.c
> > +++ b/ofproto/ofproto-dpif-monitor.c
> > @@ -63,7 +63,6 @@ static pthread_t monitor_tid;
> >  /* True if the monitor thread is running. */
> >  static bool monitor_running;
> >
> > -static struct seq *monitor_seq;
> >  static struct latch monitor_exit_latch;
> >  static struct ovs_rwlock monitor_rwlock = OVS_RWLOCK_INITIALIZER;
> >
> > @@ -149,10 +148,9 @@ mport_update(struct mport *mport, struct bfd *bfd,
> struct cfm *cfm,
> >          memcpy(mport->hw_addr, hw_addr, ETH_ADDR_LEN);
> >      }
> >      /* If bfd/cfm is added or reconfigured, move the mport on top of
> the heap
> > -     * and wakes up the monitor thread. */
> > +     * so that the monitor thread can run the mport next time it wakes
> up. */
> >      if (mport->bfd || mport->cfm) {
> >          heap_change(&monitor_heap, &mport->heap_node, LLONG_MAX);
> > -        seq_change(monitor_seq);
> >      }
> >  }
> >
> > @@ -165,7 +163,6 @@ monitor_init(void)
> >
> >      if (ovsthread_once_start(&once)) {
> >          hmap_init(&monitor_hmap);
> > -        monitor_seq = seq_create();
> >          ovsthread_once_done(&once);
> >      }
> >  }
> > @@ -177,17 +174,18 @@ monitor_main(void * args OVS_UNUSED)
> >      set_subprogram_name("monitor");
> >      VLOG_INFO("monitor thread created");
> >      while (!latch_is_set(&monitor_exit_latch)) {
> > -        uint64_t seq = seq_read(monitor_seq);
> > -
> >          monitor_run();
> >          latch_wait(&monitor_exit_latch);
> > -        seq_wait(monitor_seq, seq);
> >          poll_block();
> >      }
> >      VLOG_INFO("monitor thread terminated");
> >      return NULL;
> >  }
> >
> > +/* The monitor thread should wake up this often to ensure that newly
> added or
> > + * reconfigured monitoring ports are run in a timely manner. */
> > +#define MONITOR_INTERVAL_MSEC 100
> > +
> >  /* Checks the sending of control packets on mports that have timed out.
> >   * Sends the control packets if needed.  Executes bfd and cfm periodic
> >   * functions (run, wait) on those mports. */
> > @@ -234,7 +232,11 @@ monitor_run(void)
> >
> >      /* Waits on the earliest next wakeup time. */
> >      if (!heap_is_empty(&monitor_heap)) {
> > -
>  poll_timer_wait_until(PRIO_TO_MSEC(heap_max(&monitor_heap)->priority));
> > +        long long int next_timeout, next_mport_wakeup;
> > +
> > +        next_timeout = time_msec() + MONITOR_INTERVAL_MSEC;
> > +        next_mport_wakeup =
> PRIO_TO_MSEC(heap_max(&monitor_heap)->priority);
> > +        poll_timer_wait_until(MIN(next_timeout, next_mport_wakeup));
> >      }
> >      ovs_rwlock_unlock(&monitor_rwlock);
> >      ofpbuf_uninit(&packet);
> > diff --git a/tests/bfd.at b/tests/bfd.at
> > index ccb62b5..6583dd5 100644
> > --- a/tests/bfd.at
> > +++ b/tests/bfd.at
> > @@ -334,6 +334,7 @@ BFD_CHECK([p1], [true], [false], [none], [up], [No
> Diagnostic], [none], [up], [N
> >  # for decay_min_rx < 2000ms, the decay detection time is set to 2000ms.
> >  # this should reset the min_rx.
> >  AT_CHECK([ovs-vsctl set Interface p0 bfd:decay_min_rx=1000])
> > +ovs-appctl time/warp 100
> >  BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
> >  BFD_CHECK_RX([p0], [500ms], [300ms], [500ms])
> >
> > @@ -360,8 +361,7 @@ for i in `seq 0 9`; do ovs-appctl time/warp 500; done
> >  # Test-4 BFD decay: set min_rx to 800ms.
> >  # this should firstly reset the min_rx and then re-decay to 1000ms.
> >  AT_CHECK([ovs-vsctl set Interface p0 bfd:min_rx=800])
> > -BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
> > -BFD_CHECK_RX([p0], [800ms], [800ms], [500ms])
> > +ovs-appctl time/warp 100
> >
> >  # for the following 1600ms, there should be no decay,
> >  # since the decay detection time is set to 2000ms.
> > @@ -385,8 +385,7 @@ for i in `seq 0 9`; do ovs-appctl time/warp 500; done
> >
> >  # Test-5 BFD decay: set min_rx to 300ms and decay_min_rx to 5000ms
> together.
> >  AT_CHECK([ovs-vsctl set Interface p0 bfd:min_rx=300
> bfd:decay_min_rx=5000])
> > -BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
> > -BFD_CHECK_RX([p0], [500ms], [300ms], [500ms])
> > +ovs-appctl time/warp 100
> >
> >  # for decay_min_rx > 2000ms, the decay detection time is set to
> >  # decay_min_rx (5000ms).
> > @@ -412,9 +411,8 @@ for i in `seq 0 9`; do ovs-appctl time/warp 500; done
> >
> >  # Test-6 BFD decay: set decay_min_rx to 0 to disable bfd decay.
> >  AT_CHECK([ovs-vsctl set Interface p0 bfd:decay_min_rx=0])
> > -# min_rx is reset.
> > -BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
> > -BFD_CHECK_RX([p0], [500ms], [300ms], [500ms])
> > +ovs-appctl time/warp 100
> > +
> >  for i in `seq 0 20`
> >  do
> >      ovs-appctl time/warp 500
> > @@ -444,6 +442,8 @@ do
> >  done
> >  # reset the p1's min_tx to 500ms.
> >  AT_CHECK([ovs-vsctl set Interface p1 bfd:min_tx=500])
> > +ovs-appctl time/warp 100
> > +
> >  # since p0 has been in decay, now the RX will show 3000ms.
> >  BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
> >  BFD_CHECK_RX([p0], [3000ms], [3000ms], [500ms])
> > @@ -629,8 +629,6 @@ done
> >
> >  # reconfigure the decay_min_rx to 1000ms.
> >  AT_CHECK([ovs-vsctl set interface p0 bfd:decay_min_rx=1000])
> > -BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
> > -BFD_CHECK_RX([p0], [500ms], [300ms], [500ms])
> >
> >  # wait for 5000ms to decay.
> >  for i in `seq 0 9`; do ovs-appctl time/warp 500; done
> > diff --git a/tests/cfm.at b/tests/cfm.at
> > index 9e351d0..620e3e0 100644
> > --- a/tests/cfm.at
> > +++ b/tests/cfm.at
> > @@ -83,17 +83,19 @@ for i in `seq 0 100`; do ovs-appctl time/warp 100;
> done
> >  CFM_CHECK_EXTENDED([p0], [1], [100], [up], [up], [100ms], [2], [up])
> >  CFM_CHECK_EXTENDED([p1], [2], [100], [up], [up], [100ms], [1], [up])
> >
> > -# turn cfm on p1 off, should increment the cfm_flap_count on p1.
> > +# turn cfm on p1 off, should increment the cfm_flap_count on p0.
> >  AT_CHECK([ovs-vsctl remove interface p1 cfm_mpid 2])
> >  for i in `seq 0 10`; do ovs-appctl time/warp 100; done
> >  CFM_VSCTL_LIST_IFACE([p0], [cfm_flap_count], [cfm_flap_count      : 1])
> >  CFM_VSCTL_LIST_IFACE([p1], [cfm_flap_count], [cfm_flap_count      :
> [[]]])
> >
> > -# turn cfm on p1 on again, should increment the cfm_flap_count on p1.
> > +# turn cfm on p1 on again, should increment the cfm_flap_count on p0.
> > +# After p1 is configured, but before it is first run, it will receive
> > +# a ccm with the rdi bit set, and detect a flap as well.
> >  AT_CHECK([ovs-vsctl set interface p1 cfm_mpid=2])
> >  for i in `seq 0 10`; do ovs-appctl time/warp 100; done
> >  CFM_VSCTL_LIST_IFACE([p0], [cfm_flap_count], [cfm_flap_count      : 2])
> > -CFM_VSCTL_LIST_IFACE([p1], [cfm_flap_count], [cfm_flap_count      : 0])
> > +CFM_VSCTL_LIST_IFACE([p1], [cfm_flap_count], [cfm_flap_count      : 2])
> >
> >  OVS_VSWITCHD_STOP
> >  AT_CLEANUP
> > \ No newline at end of file
> > --
> > 1.7.9.5
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-dev/attachments/20131212/763d88b9/attachment-0003.html>


More information about the dev mailing list