From 274fda334a84253222e01b779349784ec552921b Mon Sep 17 00:00:00 2001
From: Christine Caulfield <ccaulfie@redhat.com>
Date: Fri, 21 Feb 2020 09:20:29 +0000
Subject: [PATCH] stats: Add basic schedule-miss stats to needle
In camelback (48b6894ef41e9a06ccbb696d062d86ef60dc2c4b) we have
a much more comprehensive system for recording
schedule misses because it has a 'stats' map. This is much more basic
and just writes the last event into cmap. You can still query and track
the value though.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
---
exec/main.c | 3 +++
man/cmap_keys.8 | 10 ++++++++++
2 files changed, 13 insertions(+)
diff --git a/exec/main.c b/exec/main.c
index 787d5c9..204abc8 100644
--- a/exec/main.c
+++ b/exec/main.c
@@ -852,6 +852,9 @@ static void timer_function_scheduler_timeout (void *data)
log_printf (LOGSYS_LEVEL_WARNING, "Corosync main process was not scheduled for %0.4f ms "
"(threshold is %0.4f ms). Consider token timeout increase.",
(float)tv_diff / QB_TIME_NS_IN_MSEC, (float)timeout_data->max_tv_diff / QB_TIME_NS_IN_MSEC);
+
+ icmap_set_float("runtime.schedmiss.delay", (float)tv_diff / QB_TIME_NS_IN_MSEC);
+ icmap_set_uint64("runtime.schedmiss.timestamp", qb_util_nano_from_epoch_get() / QB_TIME_NS_IN_MSEC);
}
/*
diff --git a/man/cmap_keys.8 b/man/cmap_keys.8
index b0cd721..1045c65 100644
--- a/man/cmap_keys.8
+++ b/man/cmap_keys.8
@@ -256,6 +256,16 @@ Status of the processor. Can be one of joined and left.
Config version of the member node.
.TP
+runtime.schedmiss.timestamp
+The timestamp of the last time when corosync failed to be scheduled
+for the required amount of time. The even is warned in syslog but this
+is easier to find. The time is milli-seconds since the epoch.
+
+.B
+runtime.schedmiss.delay
+The amount of time (milliseconds as a float) that corosync was delayed.
+
+.TP
resources.process.PID.*
Prefix created by applications using SAM with CMAP integration.
It contains the following keys:
--
1.8.3.1