add a new eventscript 01.reclock

    if the reclock file has been set, then this script will test that the
    reclock file can actually be accessed.
    if the file does not exist, or if the attempts to stat the file hangs,
    the node will be marked unhealthy after the third failed monitoring event
    and after the tenth failure, ctdb itself will shutdown.

diff --git a/Makefile.in b/Makefile.in
index cda8ec22d93c6ad8ee8988fe797b49ec878d6658..88f6eaef937b838342c7e102f1d1f6ad5b63136b 100755 (executable)
--- a/Makefile.in
+++ b/Makefile.in
@@ -207,6 +207,7 @@ install: all
        ${INSTALLCMD} -m 644 config/events.d/README $(DESTDIR)$(docdir)/ctdb/README.eventscripts
        ${INSTALLCMD} -m 644 doc/recovery-process.txt $(DESTDIR)$(docdir)/ctdb/recovery-process.txt
        ${INSTALLCMD} -m 755 config/events.d/00.ctdb $(DESTDIR)$(etcdir)/ctdb/events.d
+       ${INSTALLCMD} -m 755 config/events.d/01.reclock $(DESTDIR)$(etcdir)/ctdb/events.d
        ${INSTALLCMD} -m 755 config/events.d/10.interface $(DESTDIR)$(etcdir)/ctdb/events.d
        ${INSTALLCMD} -m 755 config/events.d/11.natgw $(DESTDIR)$(etcdir)/ctdb/events.d
        ${INSTALLCMD} -m 755 config/events.d/11.routing $(DESTDIR)$(etcdir)/ctdb/events.d

diff --git a/config/events.d/01.reclock b/config/events.d/01.reclock
new file mode 100755 (executable)
index 0000000..7ca71ed
--- /dev/null
+++ b/config/events.d/01.reclock
@@ -0,0 +1,58 @@
+#!/bin/sh
+# script to check accessibility to the reclock file on a node
+
+. $CTDB_BASE/functions
+loadconfig ctdb
+
+cmd="$1"
+shift
+
+PATH=/usr/bin:/bin:/usr/sbin:/sbin:$PATH
+
+# The size of this file represents the number of intervals that have
+# passed when we have tried to but failed to stat the reclock file.
+# after third failure the node becomes unhealthy
+# after the tenth failure the node we shutdown ctdbd
+RECLOCKCOUNT="$CTDB_BASE/state/reclock-fail-count"
+
+case $cmd in 
+     startup)
+       echo -n > $RECLOCKCOUNT
+       ;;
+
+      monitor)
+       echo -n 1 >> $RECLOCKCOUNT
+
+       COUNT=`ls -ln $RECLOCKCOUNT | cut -d" " -f5`
+       [ $COUNT -gt 9 ] && {
+               echo "Reclock file can not be accessed. Shutting down."
+               sleep 1
+               ctdb shutdown
+       }
+
+       RECLOCKFILE=`ctdb -Y getreclock`
+       [ -z $RECLOCKFILE ] && {
+               # we are not using a reclock file
+               echo -n > $RECLOCKCOUNT
+               exit 0
+       }
+
+       # try stat the reclock file as a background process
+       # so that we dont block in case the cluster filesystem is unavailable
+       (
+               stat $RECLOCKFILE
+               [ "$?" -eq 0 ] && {
+                       # we could stat the file, reset the counter
+                       echo -n > $RECLOCKCOUNT
+               }
+       ) >/dev/null 2>/dev/null &
+
+
+       [ $COUNT -gt 2 ] && {
+               echo "Reclock file can not be accessed. Mark node UNHEALTHY."
+               exit 1;
+       }
+       ;;
+esac
+
+exit 0

diff --git a/packaging/RPM/ctdb.spec b/packaging/RPM/ctdb.spec
index 7655e309af50b0ee77678a9d7150831653325a1b..0954f74b054e55fb3ad7b3b2e808f76e9572adef 100644 (file)
--- a/packaging/RPM/ctdb.spec
+++ b/packaging/RPM/ctdb.spec
@@ -104,6 +104,7 @@ fi
 %{_docdir}/ctdb/README.eventscripts
 %{_docdir}/ctdb/recovery-process.txt
 %{_sysconfdir}/ctdb/events.d/00.ctdb
+%{_sysconfdir}/ctdb/events.d/01.reclock
 %{_sysconfdir}/ctdb/events.d/10.interface
 %{_sysconfdir}/ctdb/events.d/11.natgw
 %{_sysconfdir}/ctdb/events.d/11.routing