Skip to content

M
MariaDB
Commit e8121e50 authored by stewart@mysql.com[stewart]'s avatar stewart@mysql.com[stewart]
Browse files
Options

[PATCH] WL#3704 mgmapi timeouts: Return sane errors for timeout in mgmapi 

In ndb_mgm_call, add checks for expired timeout in (Input|Output)Stream.
In case of timeout, we set NdbMgmHandle->last_error and return NULL.

In api calls not using ndb_mgm_call (or using it in conjunction with
own IO), they'll need to check for timeouts manually. Macros are provided
to do this.

Add ndb_mgm_disconnect_quiet(h) to disconnect without checking errors
(so we don't clobber NdbMgmHandle->last_error). This helps us provide
the *consistent* semantic that on timeout we leave the NdbMgmHandle
*disconnected*. We check for this in testMgm.

Change CHECK_REPLY in mgmapi to also check for set error in handle->last_error
This will pick up the ETIMEDOUT errors and return them to client (through
returning correct failure code for API call and setting NdbMgmHandle error).
Applications written to MGMAPI before this patch will behave as before,
and even hopefully check get_last_error and report the error back to the
end user!

Adding the last CHECK_TIMEDOUT_RET and delete in ndb_mgm_call() we
slightly change behaviour of mgmapi. Previously, if disconnect
midway through a reply, where there were only optional parameters left,
we'd get a Properties object from ndb_mgm_call() containing NULLs for
the optional parameters, leading to interesting error messages. This
enables the returning of the *real* message and actually improves the API
without breaking compatibility.

ndb_mgm_start_signallog
ndb_mgm_stop_signallog
ndb_mgm_log_signals
ndb_mgm_set_trace
ndb_mgm_insert_error
ndb_mgm_set_int64_parameter [1]
ndb_mgm_set_string_parameter [1]
ndb_mgm_purge_stale_sessions [2]
 - return error code on error during ndb_mgm_call

TODO:
ndb_mgm_report_event [2]

[1] marked for removal, unused.
[2] return codes incorrect in CHECK_HANDLE/CONNECTED. undocumented.


Server side:
 in Services (per session) add macro for injecting timeout error
 (just waiting 10 seconds before continuing... it does work!)

 We inject these errors in a number of critical places - including
 the tricky api functions that don't just use ndb_mgm_call but do
 their own thing (get_config, get_status and friends)

ATRT:
 Expand testMgm to add timout tests for API. Fully automated.
 *THEORETICALLY* timing dependent - an ultra-slow network will
 cause problems and "fake" failures... I welcome other solutions.

 Tests aren't exhaustive, but cover the generics and the tricky bits.
 Also test some calling semantics (incl disconnected on error).

 It is encouraged to add *more* mgmapi tests, not less :)

InputStream:
  Fix where timedout error is set


Index: ndb-work/storage/ndb/src/mgmapi/mgmapi.cpp
===================================================================
parent 64361ee3
Show whitespace changes
Inline Side-by-side
Showing with 500 additions and 84 deletions
storage/ndb/src/common/util/InputStream.cpp
View file @ e8121e50
......@@ -59,6 +59,7 @@ SocketInputStream::gets(char * buf, int bufLen) {
if(res == 0)
{
m_timedout= true;
buf[0]=0;
return buf;
}
......@@ -67,7 +68,6 @@ SocketInputStream::gets(char * buf, int bufLen) {
if(res == -1)
{
m_timedout= true;
return 0;
}
......
storage/ndb/src/mgmapi/mgmapi.cpp
View file @ e8121e50
......@@ -137,18 +137,41 @@ setError(NdbMgmHandle h, int error, int error_line, const char * msg, ...){
return ret; \
}
#define CHECK_REPLY(reply, ret) \
#define CHECK_REPLY(handle, reply, ret) \
if(reply == NULL) { \
if(!handle->last_error) \
SET_ERROR(handle, NDB_MGM_ILLEGAL_SERVER_REPLY, ""); \
return ret; \
}
#define DBUG_CHECK_REPLY(reply, ret) \
#define DBUG_CHECK_REPLY(handle, reply, ret) \
if (reply == NULL) { \
if(!handle->last_error) \
SET_ERROR(handle, NDB_MGM_ILLEGAL_SERVER_REPLY, ""); \
DBUG_RETURN(ret); \
}
#define CHECK_TIMEDOUT(in, out) \
if(in.timedout() || out.timedout()) \
SET_ERROR(handle, ETIMEDOUT, \
"Time out talking to management server");
#define CHECK_TIMEDOUT_RET(h, in, out, ret) \
if(in.timedout() || out.timedout()) { \
SET_ERROR(handle, ETIMEDOUT, \
"Time out talking to management server"); \
ndb_mgm_disconnect_quiet(h); \
return ret; \
}
#define DBUG_CHECK_TIMEDOUT_RET(h, in, out, ret) \
if(in.timedout() || out.timedout()) { \
SET_ERROR(handle, ETIMEDOUT, \
"Time out talking to management server"); \
ndb_mgm_disconnect_quiet(h); \
DBUG_RETURN(ret); \
}
/*****************************************************************************
* Handles
*****************************************************************************/
......@@ -375,6 +398,8 @@ ndb_mgm_call(NdbMgmHandle handle, const ParserRow<ParserDummy> *command_reply,
}
out.println("");
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
Parser_t::Context ctx;
ParserDummy session(handle->socket);
Parser_t parser(command_reply, in, true, true, true);
......@@ -382,14 +407,17 @@ ndb_mgm_call(NdbMgmHandle handle, const ParserRow<ParserDummy> *command_reply,
const Properties* p = parser.parse(ctx, session);
if (p == NULL){
if(!ndb_mgm_is_connected(handle)) {
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
DBUG_RETURN(NULL);
}
else
{
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
if(ctx.m_status==Parser_t::Eof
|| ctx.m_status==Parser_t::NoLine)
{
ndb_mgm_disconnect(handle);
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
DBUG_RETURN(NULL);
}
/**
......@@ -411,6 +439,10 @@ ndb_mgm_call(NdbMgmHandle handle, const ParserRow<ParserDummy> *command_reply,
p->print(handle->logfile, "IN: ");
}
#endif
if(p && (in.timedout() || out.timedout()))
delete p;
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
DBUG_RETURN(p);
}
......@@ -603,6 +635,22 @@ ndb_mgm_get_fd(NdbMgmHandle handle)
return handle->socket;
}
/**
* Disconnect from mgm server without error checking
* Should be used internally only.
* e.g. on timeout, we leave NdbMgmHandle disconnected
*/
extern "C"
int
ndb_mgm_disconnect_quiet(NdbMgmHandle handle)
{
NDB_CLOSE_SOCKET(handle->socket);
handle->socket = NDB_INVALID_SOCKET;
handle->connected = 0;
return 0;
}
/**
* Disconnect from a mgm server
*/
......@@ -614,11 +662,7 @@ ndb_mgm_disconnect(NdbMgmHandle handle)
CHECK_HANDLE(handle, -1);
CHECK_CONNECTED(handle, -1);
NDB_CLOSE_SOCKET(handle->socket);
handle->socket = NDB_INVALID_SOCKET;
handle->connected = 0;
return 0;
return ndb_mgm_disconnect_quiet(handle);
}
struct ndb_mgm_type_atoi
......@@ -787,18 +831,24 @@ ndb_mgm_get_status(NdbMgmHandle handle)
out.println("get status");
out.println("");
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
char buf[1024];
if(!in.gets(buf, sizeof(buf)))
{
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
SET_ERROR(handle, NDB_MGM_ILLEGAL_SERVER_REPLY, "Probably disconnected");
return NULL;
}
if(strcmp("node status\n", buf) != 0) {
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
ndbout << in.timedout() << " " << out.timedout() << buf << endl;
SET_ERROR(handle, NDB_MGM_ILLEGAL_NODE_STATUS, buf);
return NULL;
}
if(!in.gets(buf, sizeof(buf)))
{
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
SET_ERROR(handle, NDB_MGM_ILLEGAL_SERVER_REPLY, "Probably disconnected");
return NULL;
}
......@@ -807,6 +857,7 @@ ndb_mgm_get_status(NdbMgmHandle handle)
Vector<BaseString> split;
tmp.split(split, ":");
if(split.size() != 2){
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
SET_ERROR(handle, NDB_MGM_ILLEGAL_NODE_STATUS, buf);
return NULL;
}
......@@ -841,6 +892,10 @@ ndb_mgm_get_status(NdbMgmHandle handle)
if(!in.gets(buf, sizeof(buf)))
{
free(state);
if(in.timedout() || out.timedout())
SET_ERROR(handle, ETIMEDOUT,
"Time out talking to management server");
else
SET_ERROR(handle, NDB_MGM_ILLEGAL_SERVER_REPLY,
"Probably disconnected");
return NULL;
......@@ -873,6 +928,7 @@ ndb_mgm_get_status(NdbMgmHandle handle)
if(i+1 != noOfNodes){
free(state);
CHECK_TIMEDOUT_RET(handle, in, out, NULL);
SET_ERROR(handle, NDB_MGM_ILLEGAL_NODE_STATUS, "Node count mismatch");
return NULL;
}
......@@ -901,7 +957,7 @@ ndb_mgm_enter_single_user(NdbMgmHandle handle,
args.put("nodeId", nodeId);
const Properties *reply;
reply = ndb_mgm_call(handle, enter_single_reply, "enter single user", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
BaseString result;
reply->get("result", result);
......@@ -932,7 +988,7 @@ ndb_mgm_exit_single_user(NdbMgmHandle handle, struct ndb_mgm_reply* /*reply*/)
const Properties *reply;
reply = ndb_mgm_call(handle, exit_single_reply, "exit single user", 0);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
const char * buf;
reply->get("result", &buf);
......@@ -1029,7 +1085,7 @@ ndb_mgm_stop3(NdbMgmHandle handle, int no_of_nodes, const int * node_list,
reply = ndb_mgm_call(handle, stop_reply_v2, "stop all", &args);
else
reply = ndb_mgm_call(handle, stop_reply_v1, "stop all", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
if(!reply->get("stopped", &stoppedNoOfNodes)){
SET_ERROR(handle, NDB_MGM_STOP_FAILED,
......@@ -1071,7 +1127,7 @@ ndb_mgm_stop3(NdbMgmHandle handle, int no_of_nodes, const int * node_list,
else
reply = ndb_mgm_call(handle, stop_reply_v1, "stop", &args);
CHECK_REPLY(reply, stoppedNoOfNodes);
CHECK_REPLY(handle, reply, stoppedNoOfNodes);
if(!reply->get("stopped", &stoppedNoOfNodes)){
SET_ERROR(handle, NDB_MGM_STOP_FAILED,
"Could not get number of stopped nodes from mgm server");
......@@ -1174,7 +1230,7 @@ ndb_mgm_restart3(NdbMgmHandle handle, int no_of_nodes, const int * node_list,
handle->read_timeout= 5*60*1000; // 5 minutes
reply = ndb_mgm_call(handle, restart_reply_v1, "restart all", &args);
handle->read_timeout= timeout;
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
BaseString result;
reply->get("result", result);
......@@ -1301,7 +1357,7 @@ ndb_mgm_get_clusterlog_severity_filter(NdbMgmHandle handle,
Properties args;
const Properties *reply;
reply = ndb_mgm_call(handle, getinfo_reply, "get info clusterlog", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
for(unsigned int i=0; i < severity_size; i++) {
reply->get(clusterlog_severity_names[severity[i].category], &severity[i].value);
......@@ -1332,7 +1388,7 @@ ndb_mgm_get_clusterlog_severity_filter_old(NdbMgmHandle handle)
Properties args;
const Properties *reply;
reply = ndb_mgm_call(handle, getinfo_reply, "get info clusterlog", &args);
CHECK_REPLY(reply, NULL);
CHECK_REPLY(handle, reply, NULL);
for(int i=0; i < (int)NDB_MGM_EVENT_SEVERITY_ALL; i++) {
reply->get(clusterlog_severity_names[i], &enabled[i]);
......@@ -1364,7 +1420,7 @@ ndb_mgm_set_clusterlog_severity_filter(NdbMgmHandle handle,
const Properties *reply;
reply = ndb_mgm_call(handle, filter_reply, "set logfilter", &args);
CHECK_REPLY(reply, retval);
CHECK_REPLY(handle, reply, retval);
BaseString result;
reply->get("result", result);
......@@ -1458,7 +1514,7 @@ ndb_mgm_get_clusterlog_loglevel(NdbMgmHandle handle,
Properties args;
const Properties *reply;
reply = ndb_mgm_call(handle, getloglevel_reply, "get cluster loglevel", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
for(int i=0; i < loglevel_count; i++) {
reply->get(clusterlog_names[loglevel[i].category], &loglevel[i].value);
......@@ -1494,7 +1550,7 @@ ndb_mgm_get_clusterlog_loglevel_old(NdbMgmHandle handle)
Properties args;
const Properties *reply;
reply = ndb_mgm_call(handle, getloglevel_reply, "get cluster loglevel", &args);
CHECK_REPLY(reply, NULL);
CHECK_REPLY(handle, reply, NULL);
for(int i=0; i < loglevel_count; i++) {
reply->get(clusterlog_names[i], &loglevel[i]);
......@@ -1527,7 +1583,7 @@ ndb_mgm_set_clusterlog_loglevel(NdbMgmHandle handle, int nodeId,
const Properties *reply;
reply = ndb_mgm_call(handle, clusterlog_reply,
"set cluster loglevel", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
DBUG_ENTER("ndb_mgm_set_clusterlog_loglevel");
DBUG_PRINT("enter",("node=%d, category=%d, level=%d", nodeId, cat, level));
......@@ -1565,7 +1621,7 @@ ndb_mgm_set_loglevel_node(NdbMgmHandle handle, int nodeId,
args.put("level", level);
const Properties *reply;
reply = ndb_mgm_call(handle, loglevel_reply, "set loglevel", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
BaseString result;
reply->get("result", result);
......@@ -1624,7 +1680,7 @@ ndb_mgm_listen_event_internal(NdbMgmHandle handle, const int filter[],
if(reply == NULL) {
close(sockfd);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
}
delete reply;
return sockfd;
......@@ -1668,7 +1724,7 @@ ndb_mgm_dump_state(NdbMgmHandle handle, int nodeId, const int * _args,
const Properties *prop;
prop = ndb_mgm_call(handle, dump_state_reply, "dump state", &args);
CHECK_REPLY(prop, -1);
CHECK_REPLY(handle, prop, -1);
BaseString result;
prop->get("result", result);
......@@ -1705,6 +1761,7 @@ ndb_mgm_start_signallog(NdbMgmHandle handle, int nodeId,
start_signallog_reply,
"start signallog",
&args);
CHECK_REPLY(handle, prop, -1);
if(prop != NULL) {
BaseString result;
......@@ -1741,6 +1798,7 @@ ndb_mgm_stop_signallog(NdbMgmHandle handle, int nodeId,
const Properties *prop;
prop = ndb_mgm_call(handle, stop_signallog_reply, "stop signallog", &args);
CHECK_REPLY(handle, prop, -1);
if(prop != NULL) {
BaseString result;
......@@ -1805,6 +1863,7 @@ ndb_mgm_log_signals(NdbMgmHandle handle, int nodeId,
const Properties *prop;
prop = ndb_mgm_call(handle, stop_signallog_reply, "log signals", &args);
CHECK_REPLY(handle, prop, -1);
if(prop != NULL) {
BaseString result;
......@@ -1842,6 +1901,7 @@ ndb_mgm_set_trace(NdbMgmHandle handle, int nodeId, int traceNumber,
const Properties *prop;
prop = ndb_mgm_call(handle, set_trace_reply, "set trace", &args);
CHECK_REPLY(handle, prop, -1);
if(prop != NULL) {
BaseString result;
......@@ -1879,6 +1939,7 @@ ndb_mgm_insert_error(NdbMgmHandle handle, int nodeId, int errorCode,
const Properties *prop;
prop = ndb_mgm_call(handle, insert_error_reply, "insert error", &args);
CHECK_REPLY(handle, prop, -1);
if(prop != NULL) {
BaseString result;
......@@ -1919,7 +1980,7 @@ ndb_mgm_start(NdbMgmHandle handle, int no_of_nodes, const int * node_list)
Properties args;
const Properties *reply;
reply = ndb_mgm_call(handle, start_reply, "start all", &args);
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
Uint32 count = 0;
if(!reply->get("started", &count)){
......@@ -1985,7 +2046,7 @@ ndb_mgm_start_backup(NdbMgmHandle handle, int wait_completed,
reply = ndb_mgm_call(handle, start_backup_reply, "start backup", &args);
handle->read_timeout= old_timeout;
}
CHECK_REPLY(reply, -1);
CHECK_REPLY(handle, reply, -1);
BaseString result;
reply->get("result", result);
......@@ -2019,7 +2080,7 @@ ndb_mgm_abort_backup(NdbMgmHandle handle, unsigned int backupId,
const Properties *prop;
prop = ndb_mgm_call(handle, stop_backup_reply, "abort backup", &args);
CHECK_REPLY(prop, -1);
CHECK_REPLY(handle, prop, -1);
const char * buf;
prop->get("result", &buf);
......@@ -2036,7 +2097,7 @@ ndb_mgm_abort_backup(NdbMgmHandle handle, unsigned int backupId,
extern "C"
struct ndb_mgm_configuration *
ndb_mgm_get_configuration(NdbMgmHandle handle, unsigned int version) {
SET_ERROR(handle, NDB_MGM_NO_ERROR, "Executing: ndb_mgm_get_configuration");
CHECK_HANDLE(handle, 0);
CHECK_CONNECTED(handle, 0);
......@@ -2054,7 +2115,7 @@ ndb_mgm_get_configuration(NdbMgmHandle handle, unsigned int version) {
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "get config", &args);
CHECK_REPLY(prop, 0);
CHECK_REPLY(handle, prop, 0);
do {
const char * buf;
......@@ -2091,9 +2152,14 @@ ndb_mgm_get_configuration(NdbMgmHandle handle, unsigned int version) {
size_t start = 0;
do {
if((read = read_socket(handle->socket, handle->read_timeout,
&buf64[start], len-start)) == -1){
&buf64[start], len-start)) < 1){
delete[] buf64;
buf64 = 0;
if(read==0)
SET_ERROR(handle, ETIMEDOUT, "Timeout reading packed config");
else
SET_ERROR(handle, errno, "Error reading packed config");
ndb_mgm_disconnect_quiet(handle);
break;
}
start += read;
......@@ -2214,7 +2280,7 @@ ndb_mgm_alloc_nodeid(NdbMgmHandle handle, unsigned int version, int nodetype,
const Properties *prop;
prop= ndb_mgm_call(handle, reply, "get nodeid", &args);
CHECK_REPLY(prop, -1);
CHECK_REPLY(handle, prop, -1);
nodeid= -1;
do {
......@@ -2266,7 +2332,7 @@ ndb_mgm_set_int_parameter(NdbMgmHandle handle,
const Properties *prop;
prop= ndb_mgm_call(handle, reply, "set parameter", &args);
CHECK_REPLY(prop, -1);
CHECK_REPLY(handle, prop, -1);
int res= -1;
do {
......@@ -2305,6 +2371,7 @@ ndb_mgm_set_int64_parameter(NdbMgmHandle handle,
const Properties *prop;
prop= ndb_mgm_call(handle, reply, "set parameter", &args);
CHECK_REPLY(handle, prop, 0);
if(prop == NULL) {
SET_ERROR(handle, EIO, "Unable set parameter");
......@@ -2348,6 +2415,7 @@ ndb_mgm_set_string_parameter(NdbMgmHandle handle,
const Properties *prop;
prop= ndb_mgm_call(handle, reply, "set parameter", &args);
CHECK_REPLY(handle, prop, 0);
if(prop == NULL) {
SET_ERROR(handle, EIO, "Unable set parameter");
......@@ -2385,6 +2453,7 @@ ndb_mgm_purge_stale_sessions(NdbMgmHandle handle, char **purged){
const Properties *prop;
prop= ndb_mgm_call(handle, reply, "purge stale sessions", &args);
CHECK_REPLY(handle, prop, -1);
if(prop == NULL) {
SET_ERROR(handle, EIO, "Unable to purge stale sessions");
......@@ -2470,7 +2539,7 @@ ndb_mgm_set_connection_int_parameter(NdbMgmHandle handle,
const Properties *prop;
prop= ndb_mgm_call(handle, reply, "set connection parameter", &args);
DBUG_CHECK_REPLY(prop, -1);
DBUG_CHECK_REPLY(handle, prop, -1);
int res= -1;
do {
......@@ -2512,7 +2581,7 @@ ndb_mgm_get_connection_int_parameter(NdbMgmHandle handle,
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "get connection parameter", &args);
DBUG_CHECK_REPLY(prop, -3);
DBUG_CHECK_REPLY(handle, prop, -3);
int res= -1;
do {
......@@ -2574,7 +2643,7 @@ ndb_mgm_get_mgmd_nodeid(NdbMgmHandle handle)
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "get mgmd nodeid", &args);
DBUG_CHECK_REPLY(prop, 0);
DBUG_CHECK_REPLY(handle, prop, 0);
if(!prop->get("nodeid",&nodeid)){
fprintf(handle->errstream, "Unable to get value\n");
......@@ -2609,7 +2678,7 @@ int ndb_mgm_report_event(NdbMgmHandle handle, Uint32 *data, Uint32 length)
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "report event", &args);
DBUG_CHECK_REPLY(prop, -1);
DBUG_CHECK_REPLY(handle, prop, -1);
DBUG_RETURN(0);
}
......@@ -2628,6 +2697,7 @@ int ndb_mgm_end_session(NdbMgmHandle handle)
SocketInputStream in(handle->socket, handle->read_timeout);
char buf[32];
in.gets(buf, sizeof(buf));
CHECK_TIMEDOUT_RET(handle, in, s_output, -1);
DBUG_RETURN(0);
}
......@@ -2653,7 +2723,7 @@ int ndb_mgm_get_version(NdbMgmHandle handle,
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "get version", &args);
CHECK_REPLY(prop, 0);
CHECK_REPLY(handle, prop, 0);
Uint32 id;
if(!prop->get("id",&id)){
......@@ -2704,7 +2774,7 @@ ndb_mgm_get_session_id(NdbMgmHandle handle)
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "get session id", &args);
CHECK_REPLY(prop, 0);
CHECK_REPLY(handle, prop, 0);
if(!prop->get("id",&session_id)){
fprintf(handle->errstream, "Unable to get session id\n");
......@@ -2741,7 +2811,7 @@ ndb_mgm_get_session(NdbMgmHandle handle, Uint64 id,
const Properties *prop;
prop = ndb_mgm_call(handle, reply, "get session", &args);
CHECK_REPLY(prop, 0);
CHECK_REPLY(handle, prop, 0);
Uint64 r_id;
int rlen= 0;
......
storage/ndb/src/mgmapi/mgmapi_internal.h
View file @ e8121e50
......@@ -68,6 +68,8 @@ extern "C" {
*/
NDB_SOCKET_TYPE ndb_mgm_convert_to_transporter(NdbMgmHandle *handle);
int ndb_mgm_disconnect_quiet(NdbMgmHandle handle);
#ifdef __cplusplus
}
#endif
......
storage/ndb/src/mgmsrv/Services.cpp
View file @ e8121e50
......@@ -290,6 +290,8 @@ struct PurgeStruct
#define ERROR_INSERTED(x) (g_errorInsert == x || m_errorInsert == x)
#define SLEEP_ERROR_INSERTED(x) if(ERROR_INSERTED(x)){NdbSleep_SecSleep(10);}
MgmApiSession::MgmApiSession(class MgmtSrvr & mgm, NDB_SOCKET_TYPE sock, Uint64 session_id)
: SocketServer::Session(sock), m_mgmsrv(mgm)
{
......@@ -600,12 +602,22 @@ MgmApiSession::getConfig(Parser_t::Context &,
char *tmp_str = (char *) malloc(base64_needed_encoded_length(src.length()));
(void) base64_encode(src.get_data(), src.length(), tmp_str);
SLEEP_ERROR_INSERTED(1);
m_output->println("get config reply");
m_output->println("result: Ok");
m_output->println("Content-Length: %d", strlen(tmp_str));
m_output->println("Content-Type: ndbconfig/octet-stream");
SLEEP_ERROR_INSERTED(2);
m_output->println("Content-Transfer-Encoding: base64");
m_output->println("");
if(ERROR_INSERTED(3))
{
int l= strlen(tmp_str);
tmp_str[l/2]='\0';
m_output->println(tmp_str);
NdbSleep_SecSleep(10);
}
m_output->println(tmp_str);
free(tmp_str);
......@@ -748,6 +760,7 @@ MgmApiSession::endSession(Parser<MgmApiSession>::Context &,
m_allocated_resources= new MgmtSrvr::Allocated_resources(m_mgmsrv);
SLEEP_ERROR_INSERTED(1);
m_output->println("end session reply");
}
......@@ -998,12 +1011,16 @@ MgmApiSession::getStatus(Parser<MgmApiSession>::Context &,
while(m_mgmsrv.getNextNodeId(&nodeId, NDB_MGM_NODE_TYPE_MGM)){
noOfNodes++;
}
SLEEP_ERROR_INSERTED(1);
m_output->println("node status");
SLEEP_ERROR_INSERTED(2);
m_output->println("nodes: %d", noOfNodes);
SLEEP_ERROR_INSERTED(3);
printNodeStatus(m_output, m_mgmsrv, NDB_MGM_NODE_TYPE_NDB);
printNodeStatus(m_output, m_mgmsrv, NDB_MGM_NODE_TYPE_MGM);
SLEEP_ERROR_INSERTED(4);
printNodeStatus(m_output, m_mgmsrv, NDB_MGM_NODE_TYPE_API);
SLEEP_ERROR_INSERTED(5);
nodeId = 0;
......@@ -1118,8 +1135,10 @@ MgmApiSession::enterSingleUser(Parser<MgmApiSession>::Context &,
Properties const &args) {
int stopped = 0;
Uint32 nodeId = 0;
int result= 0;
args.get("nodeId", &nodeId);
int result = m_mgmsrv.enterSingleUser(&stopped, nodeId);
result = m_mgmsrv.enterSingleUser(&stopped, nodeId);
m_output->println("enter single user reply");
if(result != 0) {
m_output->println("result: %s", get_error_text(result));
......@@ -1616,12 +1635,11 @@ void
MgmApiSession::check_connection(Parser_t::Context &ctx,
const class Properties &args)
{
if(ERROR_INSERTED(1))
{
NdbSleep_SecSleep(10);
}
SLEEP_ERROR_INSERTED(1);
m_output->println("check connection reply");
SLEEP_ERROR_INSERTED(2);
m_output->println("result: Ok");
SLEEP_ERROR_INSERTED(3);
m_output->println("");
}
......@@ -1642,10 +1660,7 @@ MgmApiSession::get_mgmd_nodeid(Parser_t::Context &ctx,
{
m_output->println("get mgmd nodeid reply");
m_output->println("nodeid:%u",m_mgmsrv.getOwnNodeId());
if(ERROR_INSERTED(1))
{
NdbSleep_SecSleep(10);
}
SLEEP_ERROR_INSERTED(1);
m_output->println("");
}
......
storage/ndb/test/ndbapi/testMgm.cpp
View file @ e8121e50
......@@ -22,6 +22,7 @@
#include <random.h>
#include <mgmapi.h>
#include <mgmapi_debug.h>
#include <InputStream.hpp>
int runLoadTable(NDBT_Context* ctx, NDBT_Step* step){
......@@ -191,6 +192,8 @@ int runTestApiSession(NDBT_Context* ctx, NDBT_Step* step)
ndb_mgm_set_connectstring(h, mgm);
ndb_mgm_connect(h,0,0,0);
NdbSleep_SecSleep(1);
if(ndb_mgm_get_session(h,session_id,&sess,&slen))
{
ndbout << "Failed, session still exists" << endl;
......@@ -207,15 +210,22 @@ int runTestApiSession(NDBT_Context* ctx, NDBT_Step* step)
}
}
int runTestApiTimeout1(NDBT_Context* ctx, NDBT_Step* step)
int runTestApiTimeoutBasic(NDBT_Context* ctx, NDBT_Step* step)
{
char *mgm= ctx->getRemoteMgm();
int result= NDBT_FAILED;
int cc= 0;
int mgmd_nodeid= 0;
ndb_mgm_reply reply;
NdbMgmHandle h;
h= ndb_mgm_create_handle();
ndb_mgm_set_connectstring(h, mgm);
ndbout << "TEST timout check_connection" << endl;
for(int error_ins=1; error_ins<=3; error_ins++)
{
ndbout << "trying error " << error_ins << endl;
ndb_mgm_connect(h,0,0,0);
if(ndb_mgm_check_connection(h) < 0)
......@@ -224,10 +234,17 @@ int runTestApiTimeout1(NDBT_Context* ctx, NDBT_Step* step)
goto done;
}
ndb_mgm_reply reply;
mgmd_nodeid= ndb_mgm_get_mgmd_nodeid(h);
if(mgmd_nodeid==0)
{
ndbout << "Failed to get mgmd node id to insert error" << endl;
result= NDBT_FAILED;
goto done;
}
reply.return_code= 0;
if(ndb_mgm_insert_error(h, 3, 1, &reply)< 0)
if(ndb_mgm_insert_error(h, mgmd_nodeid, error_ins, &reply)< 0)
{
ndbout << "failed to insert error " << endl;
result= NDBT_FAILED;
......@@ -242,29 +259,229 @@ int runTestApiTimeout1(NDBT_Context* ctx, NDBT_Step* step)
else
result= NDBT_FAILED;
ndbout << "test 2" << endl;
if(ndb_mgm_is_connected(h))
{
ndbout << "FAILED: still connected" << endl;
result= NDBT_FAILED;
}
}
ndbout << "TEST get_mgmd_nodeid" << endl;
ndb_mgm_connect(h,0,0,0);
if(ndb_mgm_insert_error(h, mgmd_nodeid, 0, &reply)< 0)
{
ndbout << "failed to remove inserted error " << endl;
result= NDBT_FAILED;
goto done;
}
cc= ndb_mgm_get_mgmd_nodeid(h);
ndbout << "got node id: " << cc << endl;
if(cc==0)
result= NDBT_OK;
else
{
ndbout << "FAILED: didn't get node id" << endl;
result= NDBT_FAILED;
}
else
result= NDBT_OK;
ndbout << "TEST end_session" << endl;
ndb_mgm_connect(h,0,0,0);
if(ndb_mgm_insert_error(h, 3, 0, &reply)< 0)
if(ndb_mgm_insert_error(h, mgmd_nodeid, 1, &reply)< 0)
{
ndbout << "failed to remove inserted error " << endl;
ndbout << "FAILED: insert error 1" << endl;
result= NDBT_FAILED;
goto done;
}
cc= ndb_mgm_get_mgmd_nodeid(h);
ndbout << "got node id: " << cc << endl;
cc= ndb_mgm_end_session(h);
if(cc==0)
{
ndbout << "FAILED: success in calling end_session" << endl;
result= NDBT_FAILED;
}
else if(ndb_mgm_get_latest_error(h)!=ETIMEDOUT)
{
ndbout << "FAILED: Incorrect error code (" << ndb_mgm_get_latest_error(h)
<< " != expected " << ETIMEDOUT << ") desc: "
<< ndb_mgm_get_latest_error_desc(h)
<< " line: " << ndb_mgm_get_latest_error_line(h)
<< " msg: " << ndb_mgm_get_latest_error_msg(h)
<< endl;
result= NDBT_FAILED;
}
else
result= NDBT_OK;
if(ndb_mgm_is_connected(h))
{
ndbout << "FAILED: is still connected after error" << endl;
result= NDBT_FAILED;
}
done:
ndb_mgm_disconnect(h);
ndb_mgm_destroy_handle(&h);
return result;
}
int runTestApiGetStatusTimeout(NDBT_Context* ctx, NDBT_Step* step)
{
char *mgm= ctx->getRemoteMgm();
int result= NDBT_OK;
int cc= 0;
int mgmd_nodeid= 0;
NdbMgmHandle h;
h= ndb_mgm_create_handle();
ndb_mgm_set_connectstring(h, mgm);
for(int error_ins=0; error_ins<=5; error_ins++)
{
ndb_mgm_connect(h,0,0,0);
if(ndb_mgm_check_connection(h) < 0)
{
result= NDBT_FAILED;
goto done;
}
mgmd_nodeid= ndb_mgm_get_mgmd_nodeid(h);
if(mgmd_nodeid==0)
{
ndbout << "Failed to get mgmd node id to insert error" << endl;
result= NDBT_FAILED;
goto done;
}
ndb_mgm_reply reply;
reply.return_code= 0;
if(ndb_mgm_insert_error(h, mgmd_nodeid, error_ins, &reply)< 0)
{
ndbout << "failed to insert error " << error_ins << endl;
result= NDBT_FAILED;
}
ndbout << "trying error: " << error_ins << endl;
ndb_mgm_set_timeout(h,2500);
struct ndb_mgm_cluster_state *cl= ndb_mgm_get_status(h);
if(cl!=NULL)
free(cl);
/*
* For whatever strange reason,
* get_status is okay with not having the last enter there.
* instead of "fixing" the api, let's have a special case
* so we don't break any behaviour
*/
if(error_ins!=0 && error_ins!=5 && cl!=NULL)
{
ndbout << "FAILED: got a ndb_mgm_cluster_state back" << endl;
result= NDBT_FAILED;
}
if(error_ins!=0 && error_ins!=5 && ndb_mgm_is_connected(h))
{
ndbout << "FAILED: is still connected after error" << endl;
result= NDBT_FAILED;
}
if(error_ins!=0 && error_ins!=5 && ndb_mgm_get_latest_error(h)!=ETIMEDOUT)
{
ndbout << "FAILED: Incorrect error code (" << ndb_mgm_get_latest_error(h)
<< " != expected " << ETIMEDOUT << ") desc: "
<< ndb_mgm_get_latest_error_desc(h)
<< " line: " << ndb_mgm_get_latest_error_line(h)
<< " msg: " << ndb_mgm_get_latest_error_msg(h)
<< endl;
result= NDBT_FAILED;
}
}
done:
ndb_mgm_disconnect(h);
ndb_mgm_destroy_handle(&h);
return result;
}
int runTestMgmApiGetConfigTimeout(NDBT_Context* ctx, NDBT_Step* step)
{
char *mgm= ctx->getRemoteMgm();
int result= NDBT_OK;
int mgmd_nodeid= 0;
NdbMgmHandle h;
h= ndb_mgm_create_handle();
ndb_mgm_set_connectstring(h, mgm);
for(int error_ins=0; error_ins<=3; error_ins++)
{
ndb_mgm_connect(h,0,0,0);
if(ndb_mgm_check_connection(h) < 0)
{
result= NDBT_FAILED;
goto done;
}
mgmd_nodeid= ndb_mgm_get_mgmd_nodeid(h);
if(mgmd_nodeid==0)
{
ndbout << "Failed to get mgmd node id to insert error" << endl;
result= NDBT_FAILED;
goto done;
}
ndb_mgm_reply reply;
reply.return_code= 0;
if(ndb_mgm_insert_error(h, mgmd_nodeid, error_ins, &reply)< 0)
{
ndbout << "failed to insert error " << error_ins << endl;
result= NDBT_FAILED;
}
ndbout << "trying error: " << error_ins << endl;
ndb_mgm_set_timeout(h,2500);
struct ndb_mgm_configuration *c= ndb_mgm_get_configuration(h,0);
if(c!=NULL)
free(c);
if(error_ins!=0 && c!=NULL)
{
ndbout << "FAILED: got a ndb_mgm_configuration back" << endl;
result= NDBT_FAILED;
}
if(error_ins!=0 && ndb_mgm_is_connected(h))
{
ndbout << "FAILED: is still connected after error" << endl;
result= NDBT_FAILED;
}
if(error_ins!=0 && ndb_mgm_get_latest_error(h)!=ETIMEDOUT)
{
ndbout << "FAILED: Incorrect error code (" << ndb_mgm_get_latest_error(h)
<< " != expected " << ETIMEDOUT << ") desc: "
<< ndb_mgm_get_latest_error_desc(h)
<< " line: " << ndb_mgm_get_latest_error_line(h)
<< " msg: " << ndb_mgm_get_latest_error_msg(h)
<< endl;
result= NDBT_FAILED;
}
}
done:
ndb_mgm_disconnect(h);
ndb_mgm_destroy_handle(&h);
......@@ -272,6 +489,103 @@ int runTestApiTimeout1(NDBT_Context* ctx, NDBT_Step* step)
return result;
}
int runTestMgmApiEventTimeout(NDBT_Context* ctx, NDBT_Step* step)
{
char *mgm= ctx->getRemoteMgm();
int result= NDBT_OK;
int mgmd_nodeid= 0;
NdbMgmHandle h;
h= ndb_mgm_create_handle();
ndb_mgm_set_connectstring(h, mgm);
for(int error_ins=0; error_ins<=3; error_ins++)
{
ndb_mgm_connect(h,0,0,0);
if(ndb_mgm_check_connection(h) < 0)
{
result= NDBT_FAILED;
goto done;
}
mgmd_nodeid= ndb_mgm_get_mgmd_nodeid(h);
if(mgmd_nodeid==0)
{
ndbout << "Failed to get mgmd node id to insert error" << endl;
result= NDBT_FAILED;
goto done;
}
ndb_mgm_reply reply;
reply.return_code= 0;
if(ndb_mgm_insert_error(h, mgmd_nodeid, error_ins, &reply)< 0)
{
ndbout << "failed to insert error " << error_ins << endl;
result= NDBT_FAILED;
}
ndbout << "trying error: " << error_ins << endl;
ndb_mgm_set_timeout(h,2500);
int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_BACKUP,
1, NDB_MGM_EVENT_CATEGORY_STARTUP,
0 };
int fd= ndb_mgm_listen_event(h, filter);
if(fd==NDB_INVALID_SOCKET)
{
ndbout << "FAILED: could not listen to event" << endl;
result= NDBT_FAILED;
}
char *tmp= 0;
char buf[512];
SocketInputStream in(fd,20000);
do {
if((tmp = in.gets(buf, sizeof(buf))))
{
const char ping_token[]="<PING>";
if(memcmp(ping_token,tmp,sizeof(ping_token)-1))
if(tmp && strlen(tmp))
ndbout << tmp;
}
else
{
if(in.timedout())
{
ndbout << "TIMED OUT READING EVENT" << endl;
break;
}
}
} while(true);
if(error_ins!=0 && ndb_mgm_is_connected(h))
{
ndbout << "FAILED: is still connected after error" << endl;
result= NDBT_FAILED;
}
if(error_ins!=0 && ndb_mgm_get_latest_error(h)!=ETIMEDOUT)
{
ndbout << "FAILED: Incorrect error code (" << ndb_mgm_get_latest_error(h)
<< " != expected " << ETIMEDOUT << ") desc: "
<< ndb_mgm_get_latest_error_desc(h)
<< " line: " << ndb_mgm_get_latest_error_line(h)
<< " msg: " << ndb_mgm_get_latest_error_msg(h)
<< endl;
result= NDBT_FAILED;
}
}
done:
ndb_mgm_disconnect(h);
ndb_mgm_destroy_handle(&h);
return result;
}
NDBT_TESTSUITE(testMgm);
TESTCASE("SingleUserMode",
......@@ -284,9 +598,24 @@ TESTCASE("ApiSessionFailure",
INITIALIZER(runTestApiSession);
}
TESTCASE("ApiTimeout1",
"Test timeout for MGMAPI"){
INITIALIZER(runTestApiTimeout1);
TESTCASE("ApiTimeoutBasic",
"Basic timeout tests for MGMAPI"){
INITIALIZER(runTestApiTimeoutBasic);
}
TESTCASE("ApiGetStatusTimeout",
"Test timeout for MGMAPI getStatus"){
INITIALIZER(runTestApiGetStatusTimeout);
}
TESTCASE("ApiGetConfigTimeout",
"Test timeouts for mgmapi get_configuration"){
INITIALIZER(runTestMgmApiGetConfigTimeout);
}
TESTCASE("ApiMgmEventTimeout",
"Test timeouts for mgmapi get_configuration"){
INITIALIZER(runTestMgmApiEventTimeout);
}
NDBT_TESTSUITE_END(testMgm);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7