summaryrefslogtreecommitdiffstats
path: root/utils/cluster_fail_time.tcl
diff options
context:
space:
mode:
Diffstat (limited to 'utils/cluster_fail_time.tcl')
-rw-r--r--utils/cluster_fail_time.tcl50
1 files changed, 50 insertions, 0 deletions
diff --git a/utils/cluster_fail_time.tcl b/utils/cluster_fail_time.tcl
new file mode 100644
index 0000000..8739949
--- /dev/null
+++ b/utils/cluster_fail_time.tcl
@@ -0,0 +1,50 @@
+# This simple script is used in order to estimate the average PFAIL->FAIL
+# state switch after a failure.
+
+set ::sleep_time 10 ; # How much to sleep to trigger PFAIL.
+set ::fail_port 30016 ; # Node to put in sleep.
+set ::other_port 30001 ; # Node to use to monitor the flag switch.
+
+proc avg vector {
+ set sum 0.0
+ foreach x $vector {
+ set sum [expr {$sum+$x}]
+ }
+ expr {$sum/[llength $vector]}
+}
+
+set samples {}
+while 1 {
+ exec redis-cli -p $::fail_port debug sleep $::sleep_time > /dev/null &
+
+ # Wait for fail? to appear.
+ while 1 {
+ set output [exec redis-cli -p $::other_port cluster nodes]
+ if {[string match {*fail\?*} $output]} break
+ after 100
+ }
+
+ puts "FAIL?"
+ set start [clock milliseconds]
+
+ # Wait for fail? to disappear.
+ while 1 {
+ set output [exec redis-cli -p $::other_port cluster nodes]
+ if {![string match {*fail\?*} $output]} break
+ after 100
+ }
+
+ puts "FAIL"
+ set now [clock milliseconds]
+ set elapsed [expr {$now-$start}]
+ puts $elapsed
+ lappend samples $elapsed
+
+ puts "AVG([llength $samples]): [avg $samples]"
+
+ # Wait for the instance to be available again.
+ exec redis-cli -p $::fail_port ping
+
+ # Wait for the fail flag to be cleared.
+ after 2000
+}