summaryrefslogtreecommitdiffstats
path: root/xlators/mgmt/glusterd/src/glusterd-rebalance.c
diff options
context:
space:
mode:
Diffstat (limited to 'xlators/mgmt/glusterd/src/glusterd-rebalance.c')
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-rebalance.c73
1 files changed, 55 insertions, 18 deletions
diff --git a/xlators/mgmt/glusterd/src/glusterd-rebalance.c b/xlators/mgmt/glusterd/src/glusterd-rebalance.c
index de8f39b7d..b7b974c68 100644
--- a/xlators/mgmt/glusterd/src/glusterd-rebalance.c
+++ b/xlators/mgmt/glusterd/src/glusterd-rebalance.c
@@ -42,12 +42,27 @@ glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
int count, void *myframe);
int
glusterd_defrag_start_validate (glusterd_volinfo_t *volinfo, char *op_errstr,
- size_t len)
+ size_t len, glusterd_op_t op)
{
- int ret = -1;
+ int ret = -1;
+ xlator_t *this = NULL;
+
+ this = THIS;
+ GF_ASSERT (this);
+
+ /* Check only if operation is not remove-brick */
+ if ((GD_OP_REMOVE_BRICK != op) &&
+ !gd_is_remove_brick_committed (volinfo)) {
+ gf_log (this->name, GF_LOG_DEBUG, "A remove-brick task on "
+ "volume %s is not yet committed", volinfo->volname);
+ snprintf (op_errstr, len, "A remove-brick task on volume %s is"
+ " not yet committed. Either commit or stop the "
+ "remove-brick task.", volinfo->volname);
+ goto out;
+ }
if (glusterd_is_defrag_on (volinfo)) {
- gf_log ("glusterd", GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_DEBUG,
"rebalance on volume %s already started",
volinfo->volname);
snprintf (op_errstr, len, "Rebalance on %s is already started",
@@ -57,7 +72,7 @@ glusterd_defrag_start_validate (glusterd_volinfo_t *volinfo, char *op_errstr,
if (glusterd_is_rb_started (volinfo) ||
glusterd_is_rb_paused (volinfo)) {
- gf_log ("glusterd", GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_DEBUG,
"Rebalance failed as replace brick is in progress on volume %s",
volinfo->volname);
snprintf (op_errstr, len, "Rebalance failed as replace brick is in progress on "
@@ -66,13 +81,14 @@ glusterd_defrag_start_validate (glusterd_volinfo_t *volinfo, char *op_errstr,
}
ret = 0;
out:
- gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+ gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
return ret;
}
+
int32_t
-glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
- rpc_clnt_event_t event, void *data)
+__glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data)
{
glusterd_volinfo_t *volinfo = NULL;
glusterd_defrag_info_t *defrag = NULL;
@@ -130,8 +146,6 @@ glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
GF_DEFRAG_STATUS_STARTED) {
volinfo->rebal.defrag_status =
GF_DEFRAG_STATUS_FAILED;
- } else {
- volinfo->rebal.defrag_cmd = 0;
}
}
@@ -160,6 +174,14 @@ glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
return ret;
}
+int32_t
+glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data)
+{
+ return glusterd_big_locked_notify (rpc, mydata, event,
+ data, __glusterd_defrag_notify);
+}
+
int
glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
size_t len, int cmd, defrag_cbk_fn_t cbk,
@@ -181,7 +203,7 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
GF_ASSERT (volinfo);
GF_ASSERT (op_errstr);
- ret = glusterd_defrag_start_validate (volinfo, op_errstr, len);
+ ret = glusterd_defrag_start_validate (volinfo, op_errstr, len, op);
if (ret)
goto out;
if (!volinfo->rebal.defrag)
@@ -195,6 +217,7 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
defrag->cmd = cmd;
+ volinfo->rebal.defrag_cmd = cmd;
volinfo->rebal.op = op;
LOCK_INIT (&defrag->lock);
@@ -268,14 +291,16 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
* default timeout of 30mins used for unreliable network connections is
* too long for unix domain socket connections.
*/
- ret = rpc_clnt_transport_unix_options_build (&options, sockfile, 600);
+ ret = rpc_transport_unix_options_build (&options, sockfile, 600);
if (ret) {
gf_log (THIS->name, GF_LOG_ERROR, "Unix options build failed");
goto out;
}
+ synclock_unlock (&priv->big_lock);
ret = glusterd_rpc_create (&defrag->rpc, options,
glusterd_defrag_notify, volinfo);
+ synclock_lock (&priv->big_lock);
if (ret) {
gf_log (THIS->name, GF_LOG_ERROR, "RPC create failed");
goto out;
@@ -320,14 +345,16 @@ glusterd_rebalance_rpc_create (glusterd_volinfo_t *volinfo,
* default timeout of 30mins used for unreliable network connections is
* too long for unix domain socket connections.
*/
- ret = rpc_clnt_transport_unix_options_build (&options, sockfile, 600);
+ ret = rpc_transport_unix_options_build (&options, sockfile, 600);
if (ret) {
gf_log (THIS->name, GF_LOG_ERROR, "Unix options build failed");
goto out;
}
+ synclock_unlock (&priv->big_lock);
ret = glusterd_rpc_create (&defrag->rpc, options,
glusterd_defrag_notify, volinfo);
+ synclock_lock (&priv->big_lock);
if (ret) {
gf_log (THIS->name, GF_LOG_ERROR, "RPC create failed");
goto out;
@@ -376,7 +403,7 @@ out:
}
int
-glusterd_handle_defrag_volume (rpcsvc_request_t *req)
+__glusterd_handle_defrag_volume (rpcsvc_request_t *req)
{
int32_t ret = -1;
gf_cli_req cli_req = {{0,}};
@@ -461,6 +488,12 @@ out:
return 0;
}
+int
+glusterd_handle_defrag_volume (rpcsvc_request_t *req)
+{
+ return glusterd_big_locked_handler (req, __glusterd_handle_defrag_volume);
+}
+
int
glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr)
@@ -499,7 +532,7 @@ glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr)
case GF_DEFRAG_CMD_START:
case GF_DEFRAG_CMD_START_LAYOUT_FIX:
case GF_DEFRAG_CMD_START_FORCE:
- if (is_origin_glusterd ()) {
+ if (is_origin_glusterd (dict)) {
op_ctx = glusterd_op_get_ctx ();
if (!op_ctx) {
ret = -1;
@@ -525,8 +558,9 @@ glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr)
ret = 0;
}
}
- ret = glusterd_defrag_start_validate (volinfo,
- msg, sizeof (msg));
+ ret = glusterd_defrag_start_validate (volinfo, msg,
+ sizeof (msg),
+ GD_OP_REBALANCE);
if (ret) {
gf_log (this->name, GF_LOG_DEBUG,
"start validate failed");
@@ -629,15 +663,18 @@ glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
ret = 0;
} else {
uuid_parse (task_id_str, volinfo->rebal.rebalance_id) ;
+ volinfo->rebal.op = GD_OP_REBALANCE;
}
ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg),
cmd, NULL, GD_OP_REBALANCE);
break;
case GF_DEFRAG_CMD_STOP:
- /* Clear task-id only on explicitly stopping the
- * rebalance process.
+ /* Clear task-id only on explicitly stopping rebalance.
+ * Also clear the stored operation, so it doesn't cause trouble
+ * with future rebalance/remove-brick starts
*/
uuid_clear (volinfo->rebal.rebalance_id);
+ volinfo->rebal.op = GD_OP_NONE;
/* Fall back to the old volume file in case of decommission*/
list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks,
generated by cgit (git 2.34.1) at 2026-01-04 11:35:57 +0000