From 422ef6a2018ebf9d6765e1f2965778f42c6a9d9c Mon Sep 17 00:00:00 2001

From: Damien Ciabrini <dciabrin@redhat.com>

Date: Tue, 15 Mar 2016 18:45:13 +0100

Subject: [PATCH] galera: don't bootstrap from a node with no grastate.dat when

 possible

---

 heartbeat/README.galera |  9 ++++-----

 heartbeat/galera        | 36 ++++++++++++++++++++++--------------

 2 files changed, 26 insertions(+), 19 deletions(-)

diff --git a/heartbeat/galera b/heartbeat/galera

index 72add3c..e4495be 100755

--- a/heartbeat/galera

+++ b/heartbeat/galera

@@ -276,20 +276,20 @@ is_bootstrap()

 }

-set_heuristic_recovered()

+set_no_grastate()

 {

-    ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-heuristic-recovered" -v "true"

+    ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -v "true"

 }

-clear_heuristic_recovered()

+clear_no_grastate()

 {

-    ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-heuristic-recovered" -D

+    ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -D

 }

-is_heuristic_recovered()

+is_no_grastate()

 {

     local node=$1

-    ${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-heuristic-recovered" -Q 2>/dev/null

+    ${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -Q 2>/dev/null

 }

 clear_last_commit()

@@ -419,7 +419,7 @@ detect_first_master()

     # avoid selecting a recovered node as bootstrap if possible

     for node in $(echo "$OCF_RESKEY_wsrep_cluster_address" | sed 's/gcomm:\/\///g' | tr -d ' ' | tr -s ',' ' '); do

-        if is_heuristic_recovered $node; then

+        if is_no_grastate $node; then

             nodes_recovered="$nodes_recovered $node"

         else

             nodes="$nodes $node"

@@ -473,6 +473,12 @@ detect_last_commit()

         local tmp=$(mktemp)

         local tmperr=$(mktemp)

+        # if we pass here because grastate.dat doesn't exist,

+        # try not to bootstrap from this node if possible

+        if [ ! -f ${OCF_RESKEY_datadir}/grastate.dat ]; then

+            set_no_grastate

+        fi

+

         ocf_log info "now attempting to detect last commit version using 'mysqld_safe --wsrep-recover'"

         ${OCF_RESKEY_binary} $recover_args --wsrep-recover > $tmp 2> $tmperr

@@ -496,8 +502,8 @@ detect_last_commit()

                 if [ ! -z "$last_commit" ]; then

                     ocf_log warn "State recovered. force SST at next restart for full resynchronization"

                     rm -f ${OCF_RESKEY_datadir}/grastate.dat

-                    # try not to use this node if bootstrap is needed

-                    set_heuristic_recovered

+                    # try not to bootstrap from this node if possible

+                    set_no_grastate

                 fi

             fi

         fi

@@ -582,17 +588,17 @@ galera_promote()

     if ocf_is_true $bootstrap; then

         promote_everyone

         clear_bootstrap_node

-        # clear attribute heuristic-recovered. if last shutdown was

+        # clear attribute no-grastate. if last shutdown was

         # not clean, we cannot be extra-cautious by requesting a SST

         # since this is the bootstrap node

-        clear_heuristic_recovered

+        clear_no_grastate

         ocf_log info "Bootstrap complete, promoting the rest of the galera instances."

     else

         # if this is not the bootstrap node, make sure this instance

         # syncs with the rest of the cluster before promotion returns.

         wait_for_sync

-        # sync is done, clear info about last recovery

-        clear_heuristic_recovered

+        # sync is done, clear info about last startup

+        clear_no_grastate

     fi

     ocf_log info "Galera started"

@@ -611,6 +617,7 @@ galera_demote()

     # if this node was previously a bootstrap node, that is no longer the case.

     clear_bootstrap_node

     clear_last_commit

+    clear_no_grastate

     # record last commit for next promotion

     detect_last_commit

@@ -722,6 +729,7 @@ galera_stop()

     clear_last_commit

     clear_master_score

     clear_bootstrap_node

+    clear_no_grastate

     return $rc

 }