The branch, master has been updated
       via  49791db7dc74cffd7e88bd73091590cdc1909328 (commit)
       via  4340263b219d75c39f8de22abe3f6f1c1ee63ea2 (commit)
      from  02b62482164a3c69715949074feb7f191a29d534 (commit)

http://gitweb.samba.org/?p=ctdb.git;a=shortlog;h=master


- Log -----------------------------------------------------------------
commit 49791db7dc74cffd7e88bd73091590cdc1909328
Author: Ronnie Sahlberg <ronniesahlb...@gmail.com>
Date:   Tue Feb 28 06:58:59 2012 +1100

    Add a tunable variable to control how long we defer after a ctdb addip 
until we force a rebalance and try to failback addresses onto this node
    Have it default to 300 seconds.

commit 4340263b219d75c39f8de22abe3f6f1c1ee63ea2
Author: Ronnie Sahlberg <ronniesahlb...@gmail.com>
Date:   Tue Feb 28 06:56:04 2012 +1100

    When adding ips to nodes, set up a deferred rebalance for the whole node to 
trigger after 60 seconds in case the normal ipreallocated is not sufficient to 
trigger rebalance.

-----------------------------------------------------------------------

Summary of changes:
 include/ctdb_private.h  |    2 +
 include/ctdb_protocol.h |    5 ++++
 server/ctdb_recoverd.c  |   56 +++++++++++++++++++++++++++++++++++++++++++++++
 server/ctdb_takeover.c  |   40 +++++++++++++++++++++++++++++++++
 server/ctdb_tunables.c  |    3 +-
 tools/ctdb.c            |   52 ++++++++++++++++++++++++++++++++++---------
 6 files changed, 146 insertions(+), 12 deletions(-)


Changeset truncated at 500 lines:

diff --git a/include/ctdb_private.h b/include/ctdb_private.h
index 4b7e675..3039953 100644
--- a/include/ctdb_private.h
+++ b/include/ctdb_private.h
@@ -124,6 +124,7 @@ struct ctdb_tunable {
        uint32_t lcp2_public_ip_assignment;
        uint32_t allow_client_db_attach;
        uint32_t recover_pdb_by_seqnum;
+       uint32_t deferred_rebalance_on_node_add;
 };
 
 /*
@@ -1132,6 +1133,7 @@ int ctdb_set_single_public_ip(struct ctdb_context *ctdb,
 int ctdb_set_event_script(struct ctdb_context *ctdb, const char *script);
 int ctdb_set_event_script_dir(struct ctdb_context *ctdb, const char 
*script_dir);
 int ctdb_set_notification_script(struct ctdb_context *ctdb, const char 
*script);
+void lcp2_forcerebalance(struct ctdb_context *ctdb, uint32_t pnn);
 int ctdb_takeover_run(struct ctdb_context *ctdb, struct ctdb_node_map 
*nodemap);
 
 int32_t ctdb_control_tcp_client(struct ctdb_context *ctdb, uint32_t client_id, 
diff --git a/include/ctdb_protocol.h b/include/ctdb_protocol.h
index 4cf2789..3e466d8 100644
--- a/include/ctdb_protocol.h
+++ b/include/ctdb_protocol.h
@@ -130,6 +130,11 @@ struct ctdb_call_info {
  */
 #define CTDB_SRVID_TAKEOVER_RUN 0xFB00000000000000LL
 
+/* request recovery daemon to rebalance ips for a node.
+   input is uint32_t for the node id.
+*/
+#define CTDB_SRVID_REBALANCE_NODE 0xFB01000000000000LL
+
 /* A message id to ask the recovery daemon to temporarily disable the
    public ip checks
 */
diff --git a/server/ctdb_recoverd.c b/server/ctdb_recoverd.c
index b5a6d63..642679b 100644
--- a/server/ctdb_recoverd.c
+++ b/server/ctdb_recoverd.c
@@ -65,6 +65,7 @@ struct ctdb_recoverd {
        struct ip_reallocate_list *reallocate_callers;
        TALLOC_CTX *ip_check_disable_ctx;
        struct ctdb_control_get_ifaces *ifaces;
+       TALLOC_CTX *deferred_rebalance_ctx;
 };
 
 #define CONTROL_TIMEOUT() timeval_current_ofs(ctdb->tunable.recover_timeout, 0)
@@ -2068,6 +2069,57 @@ static void reenable_ip_check(struct event_context *ev, 
struct timed_event *te,
 }
 
 
+static void ctdb_rebalance_timeout(struct event_context *ev, struct 
timed_event *te, 
+                                 struct timeval t, void *p)
+{
+       struct ctdb_recoverd *rec = talloc_get_type(p, struct ctdb_recoverd);
+       struct ctdb_context *ctdb = rec->ctdb;
+       int ret;
+
+       DEBUG(DEBUG_NOTICE,("Rebalance all nodes that have had ip assignment 
changes.\n"));
+
+       ret = ctdb_takeover_run(ctdb, rec->nodemap);
+       if (ret != 0) {
+               DEBUG(DEBUG_ERR, (__location__ " Unable to setup public 
takeover addresses. ctdb_takeover_run() failed.\n"));
+               rec->need_takeover_run = true;
+       }
+
+       talloc_free(rec->deferred_rebalance_ctx);
+       rec->deferred_rebalance_ctx = NULL;
+}
+
+       
+static void recd_node_rebalance_handler(struct ctdb_context *ctdb, uint64_t 
srvid, 
+                            TDB_DATA data, void *private_data)
+{
+       uint32_t pnn;
+       struct ctdb_recoverd *rec = talloc_get_type(private_data, struct 
ctdb_recoverd);
+
+       if (data.dsize != sizeof(uint32_t)) {
+               DEBUG(DEBUG_ERR,(__location__ " Incorrect size of node 
rebalance message. Was %zd but expected %zd bytes\n", data.dsize, 
sizeof(uint32_t)));
+               return;
+       }
+
+       if (ctdb->tunable.deferred_rebalance_on_node_add == 0) {
+               return;
+       }
+
+       pnn = *(uint32_t *)&data.dptr[0];
+
+       lcp2_forcerebalance(ctdb, pnn);
+       DEBUG(DEBUG_NOTICE,("Received message to perform node rebalancing for 
node %d\n", pnn));
+
+       if (rec->deferred_rebalance_ctx != NULL) {
+               talloc_free(rec->deferred_rebalance_ctx);
+       }
+       rec->deferred_rebalance_ctx = talloc_new(rec);
+       event_add_timed(ctdb->ev, rec->deferred_rebalance_ctx, 
+                       
timeval_current_ofs(ctdb->tunable.deferred_rebalance_on_node_add, 0),
+                       ctdb_rebalance_timeout, rec);
+}
+
+
+
 static void recd_update_ip_handler(struct ctdb_context *ctdb, uint64_t srvid, 
                             TDB_DATA data, void *private_data)
 {
@@ -3623,6 +3675,10 @@ static void monitor_cluster(struct ctdb_context *ctdb)
        /* register a message port for updating the recovery daemons node 
assignment for an ip */
        ctdb_client_set_message_handler(ctdb, CTDB_SRVID_RECD_UPDATE_IP, 
recd_update_ip_handler, rec);
 
+       /* register a message port for forcing a rebalance of a node next
+          reallocation */
+       ctdb_client_set_message_handler(ctdb, CTDB_SRVID_REBALANCE_NODE, 
recd_node_rebalance_handler, rec);
+
        for (;;) {
                TALLOC_CTX *mem_ctx = talloc_new(ctdb);
                struct timeval start;
diff --git a/server/ctdb_takeover.c b/server/ctdb_takeover.c
index c91d2f7..bff22a8 100644
--- a/server/ctdb_takeover.c
+++ b/server/ctdb_takeover.c
@@ -1531,6 +1531,32 @@ static bool basic_failback(struct ctdb_context *ctdb,
        return false;
 }
 
+struct ctdb_rebalancenodes {
+       struct ctdb_rebalancenodes *next;
+       uint32_t pnn;
+};
+static struct ctdb_rebalancenodes *force_rebalance_list = NULL;
+
+
+/* set this flag to force the node to be rebalanced even if it just didnt
+   become healthy again.
+*/
+void lcp2_forcerebalance(struct ctdb_context *ctdb, uint32_t pnn)
+{
+       struct ctdb_rebalancenodes *rebalance;
+
+       for (rebalance = force_rebalance_list; rebalance; rebalance = 
rebalance->next) {
+               if (rebalance->pnn == pnn) {
+                       return;
+               }
+       }
+
+       rebalance = talloc(ctdb, struct ctdb_rebalancenodes);
+       rebalance->pnn = pnn;
+       rebalance->next = force_rebalance_list;
+       force_rebalance_list = rebalance;
+}
+
 /* Do necessary LCP2 initialisation.  Bury it in a function here so
  * that we can unit test it.
  */
@@ -1562,6 +1588,20 @@ static void lcp2_init(struct ctdb_context * tmp_ctx,
                        (*newly_healthy)[tmp_ip->pnn] = false;
                }
        }
+
+       /* 3rd step: if a node is forced to re-balance then
+          we allow failback onto the node */
+       while (force_rebalance_list != NULL) {
+               struct ctdb_rebalancenodes *next = force_rebalance_list->next;
+
+               if (force_rebalance_list->pnn <= nodemap->num) {
+                       (*newly_healthy)[force_rebalance_list->pnn] = true;
+               }
+
+               DEBUG(DEBUG_ERR,("During ipreallocation, forced rebalance of 
node %d\n", force_rebalance_list->pnn));
+               talloc_free(force_rebalance_list);
+               force_rebalance_list = next;
+       }
 }
 
 /* Allocate any unassigned addresses using the LCP2 algorithm to find
diff --git a/server/ctdb_tunables.c b/server/ctdb_tunables.c
index d7323ec..c5e15c1 100644
--- a/server/ctdb_tunables.c
+++ b/server/ctdb_tunables.c
@@ -72,7 +72,8 @@ static const struct {
        { "StatHistoryInterval",  1,  offsetof(struct ctdb_tunable, 
stat_history_interval), false },
        { "DeferredAttachTO",  120,  offsetof(struct ctdb_tunable, 
deferred_attach_timeout), false },
        { "AllowClientDBAttach", 1, offsetof(struct ctdb_tunable, 
allow_client_db_attach), false },
-       { "RecoverPDBBySeqNum",  0, offsetof(struct ctdb_tunable, 
recover_pdb_by_seqnum), false }
+       { "RecoverPDBBySeqNum",  0, offsetof(struct ctdb_tunable, 
recover_pdb_by_seqnum), false },
+       { "DeferredRebalanceOnNodeAdd", 300, offsetof(struct ctdb_tunable, 
deferred_rebalance_on_node_add) }
 };
 
 /*
diff --git a/tools/ctdb.c b/tools/ctdb.c
index a9fbddc..ad30989 100644
--- a/tools/ctdb.c
+++ b/tools/ctdb.c
@@ -1570,6 +1570,44 @@ static int control_moveip(struct ctdb_context *ctdb, int 
argc, const char **argv
        return 0;
 }
 
+static int rebalance_node(struct ctdb_context *ctdb, uint32_t pnn)
+{
+       uint32_t recmaster;
+       TDB_DATA data;
+
+       if (ctdb_ctrl_getrecmaster(ctdb, ctdb, TIMELIMIT(), pnn, &recmaster) != 
0) {
+               DEBUG(DEBUG_ERR, ("Unable to get recmaster from node %u\n", 
pnn));
+               return -1;
+       }
+
+       data.dptr  = (uint8_t *)&pnn;
+       data.dsize = sizeof(uint32_t);
+       if (ctdb_client_send_message(ctdb, recmaster, 
CTDB_SRVID_REBALANCE_NODE, data) != 0) {
+               DEBUG(DEBUG_ERR,("Failed to send message to force node 
reallocation\n"));
+               return -1;
+       }
+
+       return 0;
+}
+
+
+/*
+  rebalance a node by setting it to allow failback and triggering a
+  takeover run
+ */
+static int control_rebalancenode(struct ctdb_context *ctdb, int argc, const 
char **argv)
+{
+       switch (options.pnn) {
+       case CTDB_BROADCAST_ALL:
+       case CTDB_CURRENT_NODE:
+               DEBUG(DEBUG_ERR,("You must specify a node number with -n <pnn> 
for the node to rebalance\n"));
+               return -1;
+       }
+
+       return rebalance_node(ctdb, options.pnn);
+}
+
+
 static int getips_store_callback(void *param, void *data)
 {
        struct ctdb_public_ip *node_ip = (struct ctdb_public_ip *)data;
@@ -1937,17 +1975,8 @@ static int control_addip(struct ctdb_context *ctdb, int 
argc, const char **argv)
                return ret;
        }
 
-       do {
-               ret = control_ipreallocate(ctdb, argc, argv);
-               if (ret != 0) {
-                       DEBUG(DEBUG_ERR, ("IP Reallocate failed on node %u. 
Wait 3 seconds and try again.\n", options.pnn));
-                       sleep(3);
-                       retries++;
-               }
-       } while (retries < 5 && ret != 0);
-       if (ret != 0) {
-               DEBUG(DEBUG_ERR, ("IP Reallocate failed on node %u. Giving 
up.\n", options.pnn));
-               talloc_free(tmp_ctx);
+       if (rebalance_node(ctdb, options.pnn) != 0) {
+               DEBUG(DEBUG_ERR,("Error when trying to rebalance node\n"));
                return ret;
        }
 
@@ -5512,6 +5541,7 @@ static const struct {
        { "readkey",         control_readkey,           true,   false,  "read 
the content off a database key", "<tdb-file> <key>" },
        { "writekey",        control_writekey,          true,   false,  "write 
to a database key", "<tdb-file> <key> <value>" },
        { "checktcpport",    control_chktcpport,        false,  true,  "check 
if a service is bound to a specific tcp port or not", "<port>" },
+       { "rebalancenode",     control_rebalancenode,   false,  false, "release 
a node by allowing it to takeover ips", "<pnn>"},
        { "getdbseqnum",     control_getdbseqnum,       false,  false, "get the 
sequence number off a database", "<dbid>" },
        { "nodestatus",      control_nodestatus,        true,   false,  "show 
and return node status" },
        { "dbstatistics",    control_dbstatistics,      false,  false, "show db 
statistics", "<db>" },


-- 
CTDB repository

Reply via email to