On Fri, Apr 20, 2012 at 02:10:18PM -0700, Ben Pfaff wrote:
> I'd always assumed that the EWMA code here was sufficient rate-limiting,
> but I actually encountered a pathological case some time ago that forced
> this rusage information to print once a second or so, which seems too
> often.
>
> Signed-off-by: Ben Pfaff <b...@nicira.com>
Here's a version of the patch generated with -w to ignore white space
changes. It shows how very little actually changed; most of the
changes are just re-indenting.
From: Ben Pfaff <b...@nicira.com>
Date: Fri, 20 Apr 2012 13:43:54 -0700
Subject: [PATCH] timeval: Rate-limit logging rusage information.
I'd always assumed that the EWMA code here was sufficient rate-limiting,
but I actually encountered a pathological case some time ago that forced
this rusage information to print once a second or so, which seems too
often.
Signed-off-by: Ben Pfaff <b...@nicira.com>
---
lib/timeval.c | 11 ++++++++---
1 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/lib/timeval.c b/lib/timeval.c
index 3f12b50..9829100 100644
--- a/lib/timeval.c
+++ b/lib/timeval.c
@@ -463,6 +463,9 @@ log_poll_interval(long long int last_wakeup)
/* Warn if we took too much time between polls: at least 50 ms and at least
* 8X the mean interval. */
if (n_samples > 10 && interval > mean_interval * 8 && interval > 50 * 16) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 3);
+
+ if (!VLOG_DROP_WARN(&rl)) {
const struct rusage *last_rusage = get_recent_rusage();
struct rusage rusage;
@@ -471,8 +474,10 @@ log_poll_interval(long long int last_wakeup)
"is over %u times the weighted mean interval %u ms "
"(%u samples)",
now - last_wakeup,
- timeval_diff_msec(&rusage.ru_utime, &last_rusage->ru_utime),
- timeval_diff_msec(&rusage.ru_stime, &last_rusage->ru_stime),
+ timeval_diff_msec(&rusage.ru_utime,
+ &last_rusage->ru_utime),
+ timeval_diff_msec(&rusage.ru_stime,
+ &last_rusage->ru_stime),
interval / mean_interval,
(mean_interval + 8) / 16, n_samples);
if (rusage.ru_minflt > last_rusage->ru_minflt
@@ -493,7 +498,7 @@ log_poll_interval(long long int last_wakeup)
rusage.ru_nvcsw - last_rusage->ru_nvcsw,
rusage.ru_nivcsw - last_rusage->ru_nivcsw);
}
-
+ }
/* Care should be taken in the value chosen for logging. Depending
* on the configuration, syslog can write changes synchronously,
* which can cause the coverage messages to take longer to log
--
1.7.2.5
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
